社群評鑑機制之求職推薦應用

i

國立交通大學

資訊管理與財務金融學系

資訊管理碩士論文

社群評鑑機制之求職推薦應用

A Social Referral Mechanism for Job Reference

Recommendation

研究生：傅渝婷

指導教授：李永銘 博士

ii

社群評鑑機制之求職推薦應用

A Social Referral Mechanism

for Job Reference Recommendation

研 究 生：傅渝婷 Student：Yu-Ting Fu

指導教授：李永銘 Advisor：Yung-Ming Li

國立交通大學

資訊管理與財務金融學系

資訊管理碩士論文

A Thesis

Submitted to Institute of Information Management College of Management

National Chiao Tung University in partial Fulfillment of the Requirements

for the Degree of Master

in

Institute of Information Management June 2013

Hsinchu, Taiwan, Republic of China

iii

社群評鑑機制之求職推薦應用

學生：傅渝婷 指導教授：李永銘 博士 國立交通大學資訊管理研究所碩士班

摘要

近年來，隨著社群平台使用者的大量增加，這類新科技的趨勢對現在人們的生活 形成巨大衝擊，也重新定義我們和他人的互動行為、增進不同社群團體的互動機會、也 讓人更容易利用群眾力量進行搜尋或尋求評價。在本次研究中，我們根據社會學家的研 究，利用新科技加速資訊的交換和傳播流程，並以求職活動當作研究的範疇。本研究即 以求職活動為例，依據協助意願和工作相關的影響力，為求職者的理想工作推薦出合適 的諮詢對象，以提供更多工作相關資訊或是進行引薦。整合了人力資源學說的研究，藉 著社群平台的技術，我們建置社群工作引薦平台，讓求職者從被動的職缺搜尋到主動的 資訊獲得，也讓社交平台的服務更加多元豐富。 關鍵詞：社群搜尋、商務社群平台、求職工作、社群推薦、社群評價

iv

A Social Referral Mechanism for Job Reference

Recommendation

Student: Yu-Ting Fu Advisor: Yung-Ming Li

Institute of Information Management

National Chiao-Tung University

ABSTRACT

Recently, with the popularity of various social media, this new trend of information technologies has impacted our lives, redefined the way we interact with each other, and facilitated the communication and influence cross different social groups, such as enhancing the power of social search and appraisal.

In this research, we mainly focus on this mystery process of information exchanges existing long ago on the base of sociology and apply this power in the field of job seeking. Considering the factors of both willingness and influence, we generate the list of proper reference candidates to desired job for job seekers to provide more job-related information or to be referrals. Integrating the knowledge of human resources management, we implement this social referral application with the support of information technologies and strive to enrich the service of social media, turning the passively searching for job seeking to actively consulting for exclusively job information.

Keywords: Social Search, Business Social Platform, Job-seeking, Social Referral, Social

v

致謝

一下子就畢業了，這兩年的時間過得真的很快，在我尚未熟悉交大的一切之前，我 又要離開這個美麗的校園了。在這兩年的時間內，第一個要感謝的人就是指導教授李永 銘老師。每當我在研究中遇到困惑的時候，老師總以自己獨特的見解指點我，給我許多 創新的觀點和解法，幫助我順利地進行一個創新服務的開發。同時也感謝口試委員劉敦 仁老師、陳柏安老師和翁頌舜老師，老師們都給予我許多建議和指教，也讓我的研究更 完整。 在論文發想和衝刺的過程中，博士班的學長姐也都給我們充足的幫助和珍貴的建議。 感謝易霖學長即使已經畢業，也願意抽空跟我討論我的論文研究，從主題選定到實驗流 程，學長都以自身的研究經驗提點，使我的研究不斷往前邁進，更趨完整，誠心感謝學 長對我們的這份關心。除了易霖學長外，也謝謝實驗室的其他學長姐，發哥學長、無尾 熊學長、智華學姊和檸瑤學姊，雖然因為時間的限制，無法和學長姐們有充足的討論， 但學長姐們都在論文實作過程上給予我適當的幫助，讓我有所成長。另一方面，也感謝 過去學長姊的幫助，因為學長姊們優秀的研究，讓我們的論文能所有借鏡，在此之上突 破，特別是小球、文翔、阿雅學長姊的研究，都對我的論文有很大的幫助。另外，我也 感謝上屆的學長們，看著你們的研究背影，也讓我對碩二的研究生活更有體悟。 同時也感謝實驗室同屆的戰友:復勛、欣宸、銘彥、Alvin。感謝復勛總願意抽時間 跟我討論我的論文內容，從架構、流程、實作都給我很多幫助和建議，一起修圖、抒發 心情，也因為有了你樂天開朗的態度，讓我的碩二生活有了更多樂趣，無論是研究或是 生活，都謝謝有你的陪伴，有你真好。欣宸，雖然你一直在打 lol 和玩手機遊戲也不簽 博，但我還是很感謝你幫我們處理了許多危機，也願意給我的論文許多意見，在最後的 衝刺時期留下來陪我改論文，也給我很多生活上的建議。Alvin 的研究進度總帶領大家 前進，也謝謝你幫助我的英文寫作。感謝銘彥在端午節的假期一起並肩作戰寫論文，也 謝謝你給大家帶來許多歡樂。此外，也謝謝實驗室的學弟妹敬媛、大天、智勝、彥丞給 我許多論文建議，幫助我們處理口試的大小事，有了你們的陪伴，實驗室有了更多歡笑

vi 和樂趣，我的研究生活也有了更多色彩，感謝你們對於我們畢業的祝福，有了你們更實 驗室更溫暖。 最後，要謝謝我的小精靈和我的家人。感謝小精靈在我忙碌求職之餘給我最直接的 幫助，和我討論論文方向、技術架構、教導我實作系統，在最關鍵的時刻全力幫助我， 成為我最強力的後盾，包容我的壞脾氣，在我焦慮難過時給我安慰，讓我沒有顧慮的往 前衝刺。沒有你的幫助，我的研究無法如此順利的完成，謝謝你。感謝我的家人支持我 完成碩士研究，關心我生活上的大小事，督促我在人生道路上往前邁進。謝謝所有在碩 士兩年生活中和我一起度過、幫助我的每個人，感謝你們，讓我有快樂的兩年時光，有 了最珍貴的學生回憶。 傅渝婷 2014 年 六月 謹致於 新竹市國立交通大學光復校區

vii

INDEX

摘要 ... iii

ABSTRACT ... iv

CHAPTER 1 INTRODUCTION ... 1

1.1Background ... 1

1.2 Motivation and Research Problems ... 3

1.3 Research Goal and Contributions ... 6

1.4Thesis Outline ... 7

CHAPTER 2 LITERATURE REVIEW ... 8

2.1 Online Recruitment ... 8

2.2 Social Media ... 8

2.3 Social Ties Analysis ... 9

2.4 Social Search ... 10

2.5 Social Support and Appraisal ... 11

CHAPTER 3 THE SYSTYEM FRAMEWORK ... 12

3.1Social Search Module ... 16

3.2 Social Appraisal Module ... 20

CHAPTER 4 EXPERIMENTS ... 32

4.1 Experiment Process ... 33

4.2 Data collection ... 35

4.3 Weight Generation ... 39

CHAPTER 5 RESULTS AND EVALUATION... 40

5.1 Accuracy of Social Referral Information ... 40

5.2 Components Weighting Determination ... 41

5.3 Recommendation Performance ... 43

5.4 Factors Performance ... 44

5.5 User’s rating ... 46

CHAPTER 6 DISCUSSION AND CONCLUSION ... 51

6.1 Research Contributions ... 51

6.2 Research Limitations ... 52

6.3 Future Works ... 54

1

CHAPTER 1 INTRODUCTION

1.1Background

Recent years, Internet has played a vital role. According to the report of International

Telecommunication Union, the number of people using Internet is going to reach 2.7 billion in the end of October 2013. Almost 40% of the world’s population is using the Internet. Internet now has become irreplaceable in our daily lives and widely applied in many ways.

For business companies, Internet has replaced the classifieds advertisement in the newspapers

and turned out to be the main channels for recruitment. According to the study in 2001 by

iLogos, it showed that 88% of the global 500 companies had a company Internet recruitment

site [1]. By 2005, 96% of all companies will use the Internet for their recruitment needs [2].

For job seekers who just graduated from colleges and feel eager to find their jobs, the job

websites are also their first choice for job-hunting by 72.7%. [3]

While most of jobs are filled through online recruitments, the human resources also

noticed that there is another effective channel, which already existed for thousands of years to

locate a job; that is by personal relationship, or we called it social network nowadays. The

survey conducted by the society for human resource management and Wall Street Journal in

2001 showed that 95% of human resources managers or job-seekers find the desired

employees or ideal jobs through the personal relationship. 61% of job recruiters and 78% of

2

As contacting as many people as possible is a highly effective way to find a job, the

social scientists precisely analyze the relationship and interaction between the job-finders and

the job information providers. Those people who are belonged to distinct social circles and

meet occasionally would bring us novel information, which includes new job offerings and

opportunities. This acquaintance relationship, which is labeled as weak tie, becomes crucial

and remarkable while considering mobility opportunities. The nature of weak ties, which

travel through different social circles, could assist job-seeker to exposes to all the information

traffic [5].

While considering the strength of weak ties, the popularity and prevalence of social

media sites nowadays can offer a perfect opportunity to practice the social theory. There are

1.26 billion people using Facebook by October 2013, which are almost half of total Internet

users, and the daily active users on average are 757 million by December 2013[6]. Social

media could gather the power of social networking and come out to be the perfect platform to

initiate social search. Social search, defined as people search or search for people via social

networks as human intermediary search, combines the basic ideas of web search mainly

based on huge amounts of database and extends it to the people (or should label as users)

involved situations[7]. It can support people to obtain information or connect people who

might help through various social circles, which can also apply in the job hunting field. The

3

and job-seekers as promotional and advertising channels. Among all the social media

platforms for job search, LinkedIn has become one of the most typical social networking

websites for people in professional occupations. As of Jun 2013, LinkedIn reports more than

259 million users across more than 200 countries [8][9]. The convenience of Internet has

conquered the boundaries of time and space constraint, which enables LinkedIn to transfer

information about occupations through distinct social circles and becomes a global company.

To sum all, since the large amount of users on the social networking sites and abundant

information generated along within those social medias, the trend of social support for

job-seeking is surging.

1.2 Motivation and Research Problems

Actually, with the convenience of various current job websites and business social

platform, there is something to be improved. First, as for the job websites, which are the main

recourses for job seekers, although they have listed lots of job openings on the websites, most

job seekers can only see the job name and very little information about that job. For example,

like Figure 1, all users can see it from that kind of websites is only the job title, the location

of that job, how many people have already applied it, and official job description.

Job users have to spend a lot of time to survey more about the job itself, such as the working

hour, real workload, and company culture about that company on the Internet randomly.

4

true information about that position or even harder to actually connect to someone working in

that company or relevant industry.

On the other hand, recently, with service supported by the famous business social

platform, LinkedIn, the job seekers could see through all the company as well as some

workers in that company. However, since LinkedIn is only for the purpose of social

networking, it lacks the consideration of real, frequent and timely interaction among people

compared to other social platforms, such as Facebook. According to the survey conducted by

The Buntin Group and Survey Sampling International, while 76% of Twitter and Facebook

users log in at least once each day, only 40% of LinkedIn user have the same habit, and even

48% of LinkedIn users only access to the website once to several times a week[10].

Undoubtedly, the less frequent users stay, the less the requests distributed on that platform

could be seen and fulfilled. Besides, it is not easy for job seekers to initiate conversation to

someone who is totally a stranger or even consul some very confidential information about

concerning job position and company. With the social relation as the mediation, in the end,

job seekers can only see the name of those employers about their dreaming companies

displaying on the website but do not have that opportunity to get closer. We may wait the Figure 1. Job Information on 104 Human Resource Banking

5

reply from someone we want to connect in the LinkedIn for days while this person may

already spend three hours lingering on the Facebook each day. Though the job seekers can

surely try to connect to anyone on LinkedIn, however it lacks of consideration the social

interactions happened in the real world.

To exploit the power of social networking in supporting job seeking, there are two problems we should solve:

1. How to benefit the job hunting activates by providing extra information about the job description through locating reference people from the social network?

First, considering the previous services provided by job websites and social platforms, what we could find through the synergy of these information is only about the job opening and some people who worked in that company. However, what we cannot obtain from those

6

combination is the social link, which here is the real social connections, happening in user’s daily live and private live, from the job seekers to someone he or she wants to contact to. Therefore, we wish to provide with more information to increase the availability through taking more factors into account.

2. How to combine the current database from the job websites and from the social network platform to discover the best consulting candidate for the job seeker?

Here we see our problem from the view of social computing. As we mentioned before, since social media is widely used among people, using this platform is perfect to start a social search. The problem we have to think about is that how to distribute the request of social search and how to find out the best candidate through all involvers. Moreover, based on what kind of information should we index and thus categorize people in order to calculate the score for ranking the candidate list on the probability to help. To resolve the problem, we need to integrate and analyze the database of both job websites and social platform.

1.3 Research Goal and Contributions

In this paper, we aim to exploit the power of social networks in improving job search. Specifically, we develop a social referral mechanism applied in the job-searching field and improve the job-matching with the power of social ties. Unlike the existing social recruiting services, which mainly focus the benefits for the recruiters and companies, we will commit to construct a social search and appraisal engine striving for the convenience of job seekers.

In this research, the main components of the proposed system include personal preference, social search and social appraisal analysis modules. First, we will first analyze what kinds of jobs users are looking for and the personal information about users as the input query. Secondly, we will ask users to distribute those requests to their own social network to collect the information we need to estimate if there are some people existing in the social circle who are related to their desired jobs. Finally, after gathering all information, we use the

7

interaction data in the social network to measure the willingness how specific person will help users or not. After those three analysis steps, users can gain a list of people they might consult about their desired job positions.

1.4Thesis Outline

The outline of the paper is organized as follows. We’ll refer to some literature review and basic concepts in Section 2. In Section 3, we present the social referral mechanism combined with social relationship analysis, personal preference analysis and job analysis. After the system framework, Section 4 describes the experiment processes and discusses the empirical results. Lastly Section 5 will cover conclusion and future research.

8

CHAPTER 2 LITERATURE REVIEW

2.1 Online Recruitment

Generally speaking, there are two channels for the company to identify and attract the potential employee: formal and informal [42]. Formal channels refer to that there are other organizations and agencies involved in the process of connecting the employees and job-seekers, such as job fairs, ads in the newspapers, personnel consultancies, online links among government centers, education institutes and online human resources agencies. The informal channels indicate it is the personal recommendation from internal employers, friends, or relatives that facilitate the process of recruiters and potential employers.

Considering so various channels with respect of recruitment, the difference and the effectiveness of them have been a main issue in the study of human resources and could be measured and estimated in distinct ways [43]. While the company enjoyed the convenience and cost-effectiveness brought by the internet recruitment sites, 33% of European companies indicates the employers recruited through online websites are more liable to leave their jobs and 44% think that it not easy to find out the very good-fit employers with the internet tools [44]. On the other hand, the employers who entered the company with personal referral would significantly work longer and also tend to accept the job offer [45]. In this research, we try to improve the effectiveness of online recruitment by integrating the idea of informal referral into the online web sites.

2.2 Social Media

With the prevalence of mobile devises and social network platform, the involvement of social media has become daily routine for people around the world. Facebook, Youtube and Baidu are the second, third and the fifth most popular websites by Feb 2014. While the activities of social actions are unveiled, many applications have been developed, no matter in the form of web applications or app, and apply in distinct fields [46], such as online dating

9

websites, knowledge transmission, finding experts [47], and fund-raising.

For the study of computer science, the vast population using the social media is an excellent opportunity to examine large-scale social theories. For example, Gilbert and Karahalios predict the tie strength and verify the completeness of dimension [48]. Choudhury, Counts and Horvitz focus on the changes of activities and emotions as to childbirth with social media interfering [49]. Burke and Kraut verify the premise that the strong ties offer better emotional support and the weak ties enable people to find a new job under the situation of unemployment [50]. Recently, there are lots of application about recruiting, as well as providing job information based on social network sharing, such as the popular business social network sites LinkedIn and Glassdoor, which has over 10 million users by 2012 [51]. The aim of this paper is to utilize the abundant data of individual’s social ties from the social media platform as well as the convenience and amount of users in order to generate a social referral application with the purpose of facilitating the job-hunting process.

2.3 Social Ties Analysis

Social network analysis has played a key role in modern research of sociology, which illustrates the connection between two individuals as “tie”. [33] Considering the concept of relationship, the intuitive notions of the “strength” of those ties would appear and thus could be calculated by different elements. Granovetter defines tie strength as a “combination of the amount of time, the emotional intensity, the intimacy (mutual confiding), and the reciprocal services which characterize the tie” [34]. Furthermore, he indicates there are two different social ties, which vary on the strength scores. The one is the “Strong Tie”, which usually occurs between trusted friends and families. The other is the “Weak Tie”, which often happens among acquaintances. [35]

Those ties impacts people’s daily lives in different ways. Reliable friends and close families can affect emotional health [36], help people suffer from stress [37], and often join

10

together to lead the groups while facing the crisis [38]. Loose friends can help a friend inspire new ideas [39], a perfect place to launch the diffusion of information, or find a job through reference [40] [41]. In this research, we utilize the function and properties of weak ties as a perfect tool to locate a social referral for job-seekers.

2.4 Social Search

For decades, while the search engine generating from the concept of digital library

project dominates the world of information retrieval [11], there are another ancient way to

acquire knowledge-“the village diagram”. Compared with the latter web search only focuses

on the individual’s seeking, the village way of query put more emphasizes on the

context-finding the right person to answer the question and also find the people connecting to

those answers [12].

Actually, this kind of concept is widely used to optimize the results of searches, such as

using collaborative social interactions [13] or social recommendation for collaborative

filtering [14] by leveraging the data in the user’s social network. Since the social acts and

social interactions could benefit the search process [15], this type of search can even be

applied in the wall of decentralized search [16]. In this search model, how to route the queries

over a social network becomes the main issues to break through [17], including considering

the factors, such as how a specific node on that node responds or not, or how relevant this

node is to the question we are searching for [18]. Although this paradigm of decentralized

11

experiment a few years ago [19], they are all bounded by the limitation that the only source of

interpersonal data collection could only be generated from the email system.[20][21]

Recently, besides the elaboration of application on the expert-finding area [22], social search

has extended to various field, such as social discover ([23]), finding new web services ([24]),

people search ([25]), and etc. In this study, we further use the concept of people search to find

the reference candidates for our desired jobs.

2.5 Social Support and Appraisal

The provision of social appraisal can be regarded as one of the important features for

social support, which is the combination of psychological and behavior functions. S.

Wasserman and K. Faust have proved that with more connections, whether in link of

friendship or interaction, it is more likely for people to influence to each other. [26] Actually,

this kind of influence play an important role when talking about social support, which is

defined as a mediating construct providing help from other people in the social network.[27]

With the help or support, or here we can say, the information from others in the social

network could be offered as the source of social appraisal to help the decision making

process.[28]

Recently, social appraisal has been widely used due to the popularity of social network

analysis. For example, a lot of researches have been conducted in the field of electronic

12

is also used in knowledge management, such as expert-finding ([31] [32]). In this research,

we use the social network relationship as the sources of social support to do the social

appraisal for the evaluating how people are willing to help you and also how people could

affect others.

CHAPTER 3 THE SYSTYEM FRAMEWORK

When it comes to job hunting, it means the processes or actions of searching for employment and the main purpose is to gain the opportunities to be interviewed by a hiring manager. Those people who seek jobs would first browse through all available sources of job information, such as job websites or newspapers, and choose those jobs which interest them. Then those job-seekers would contact their desired companies by submitting resumes or related required documents for further chances to get an interview.

13

steps: locating a job, researching employees, networking, applying, and interviewing. In this paper, we mainly pay the attention on the first three steps and design the whole mechanism to increase the possibilities to attain the interview for job seekers. Especially, for the first step to locate the job opening, we emphasize the enhancement of current job websites, which have been the main channels for job searching.

For current job websites, it can provide the job-seekers with lots of job offerings stored in the huge database collected from all kinds of companies, which improves the efficiency of doing job search as well as the variety of job opportunities and facilitates the process of locating a job. However, bounded by the base of search engine, the job websites still hold some constraints. First, in the view of the mechanism of the job websites itself, the preciseness of keyword input affects the quality of result. The misuse of searching keyword may disturb the operation of the searching process and cause less useful job information for users. Secondly, even though job sites could search all the possible job offers quickly, there is no sufficient job-related information for job-seekers to evaluate and learn more about each positions or the companies. After locating jobs, the uses still have to spend time to go through the Internet and search for any detail with regard to their desired job. Then, following by locating a job and researching, the third step mentioned above would be decisive and also be a challenge for most of job seekers; that is networking. How to discover the right reference that can affect the opportunities by considering the willingness of all the candidates within one’s social network is important for the job seekers.

In the framework of social referral mechanism, we are dedicated to resolve this problem

by combining the social network of job-seekers and the current job openings of desired

14

1. The social search engine is triggered when job seeker enters their job preferences. By this information, then we would mine the current job offerings from job websites and use the properties of job offerings as the condition of job query.

2. We could search users who are related to the job query recursively in the social network and include and update those users to the possible candidate set. Besides the candidate, the social links between those people are recorded too.

3. After locating all the possible candidates, the system will use social appraisal module to evaluate and rank all the possible referrals. The output of our system would be a list of current job openings, a ranking list of possible referral candidates about those job openings as well as the social path link to those candidates.

To meet the objective of searching for the proper candidates for reference from the

social network and rank them considering the possibility, several techniques are required. We

implement these processes by two main components in our proposed system. The system Figure 3. The process flow of social referral mechanism

15

architect is displayed as Figure 4.

1. Network Construction: In this stage, we will collect the personal and working

information about users and record the interaction data between each users.

2. Social Search Module: First we will ask user to enter their job preference and the system will discover current available offerings as the query condition. Furthermore, the system will search all the users in the social network and find out possible candidates by evaluation of job relevance.

(1) Job Preference Analysis and Opening Mining: The system will provide selections for user to choose their preference and then the system will list those job openings that match their requirement.

(2) Job Influence Analysis: The system will analyze the job information about each user to measure how relevant the user is to the job query.

3. Social Appraisal Module: Based on all the needed information, we will rank those candidates by considering both willingness and influence. Analyses about social tie strength and experience similarity are both involved. As for influence, besides the job influence, here the social influence is also considered.

(1) Referral Willingness Analysis: The system will take the social interaction and common experience between each pair of friends as the index to compute how a friend is willing to help each other.

(2) Social Influence Analysis: The system will check the social network data to know how certain use can affect his or her social network.

16

3.1 Social Search Module

The social search module measures the job relevance of each user in the social network

as the job influence to help the job seeker. Due to the nature of social network, the system

will go through the job checking process recursively.

Figure 4. Architecture of Social Referral Mechanism

17 3.1.1 Job Preference Analysis and Offer Mining

The purpose of this analysis is to search all possible job offerings in the desired industries,

companies or functions that users may be interested in. Before actually going to the search

phase, first we have to carefully describe our query condition. In this research, we refer the

common job description shown on the job website and use four properties to describe the job

shown as Figure 5. They are industry categories, company name, function name and job

grade, and each of those variables all belong to their individual set. The job could be denoted

as:

• 𝐽𝑜𝑏 = {𝐼_𝛼, 𝐶𝛽,𝐹𝛾, 𝐺𝛿}, 𝑤ℎ𝑒𝑟𝑒 𝐼𝛼𝜖𝐼, 𝐶𝛽𝜖𝐶, 𝐹𝛾𝜖𝐹, 𝐺𝛿𝜖𝐺 (1) : the set of industry category, : the set of company category,

I C

: the set of function category, : the set of distinct job grade

F G

For example, we will describe the job shown in Figure 1 in chapter 1 as this set:

1 3 5 3

{ , , , },

Job I C F G

1

I = Information Technology, C =TSMC, 3 F5= Information Technology, G3= Employee.

The stage is conducted by the following two processes:

Job Preference Analysis. The system will ask users to choose their desired job by the

Figure 5. Job Description

Job

18

input of the four variables. We will design the options of those variables by the reference of

job websites.

Opening Mining. After understanding the job seeker’s preference, the search will be

executed through all the online web sites as the database to mine the available job offerings

by the tools of web crawlers. The system will store the list of job openings in the job set for

users. For example, the job offerings which are suitable for job seeker u are:

𝑂𝑝𝑒𝑛𝑖𝑛𝑔𝑠(𝑢) = {𝐽𝑜𝑏𝑢1, 𝐽𝑜𝑏𝑢2, … , 𝐽𝑜𝑏𝑢𝑛} (2) 3.1.2 Job Influence Evaluation

In this process, our goal is to search and construct the candidate network expanded from the job seeker, or called initiator in our model. We calculate the job influence score by comparing the four indexes describing the job offering and the information of working experiences about each user in the social network. In other word, the job influence is measured as the similarity. We consider one suitable job offer at a time, denote as theJ_u. The similarity between the job offer J and the job to search _u J can be estimated as: _v

1, if ( ) ( ) ( , ) 0, otherwise u v Industry u v Industry J Industry J Sim J J     (3) 1, if ( ) ( ) ( , ) 0, otherwise u v Company u v Company J Company J Sim J J     (4) 1, if ( ) ( ) ( , ) 0, otherwise u v Function u v Function J Function J Sim J J     (5) 0.2 , 0.5 , ( ) 0.7 ,if ] 1 , v v JobGrade v v v if J is assciate if J is staff Score J

J is middle level manager if J is high level manager

        _     _    (6)

19

The values ofSim_Industry(J J_a, _b) ,Sim_Company(J J_a, _b) , and Sim_Function(J J_a, _b)present the similarity between Job a and Job b respectively. The function Industry J( _u),Company J( _u),

( _u)

Function J will return the property of the J_u respectively. If the person works in exactly the same industry, the same company or the same function that is the same as the job user u is searching for, we mark 1. Otherwise is zero. Since we only consider how the user v influence the job offer, we only user the job grade of user v.

The job influence between job of user v andJ_u:

𝐽𝐼(𝐽_𝑢,𝐽_𝑣) = 𝛼 ∗ 𝑆𝑖𝑚_{𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦}(𝐽_𝑢,𝐽_𝑣) + 𝛽 ∗ 𝑆𝑖𝑚_{𝐶𝑜𝑚𝑝𝑎𝑛𝑦}(𝐽_𝑢,𝐽_𝑣) + 𝛾 ∗ 𝑆𝑖𝑚_{𝐹𝑢𝑛𝑐𝑡𝑖𝑜𝑛}(𝐽_𝑢,𝐽_𝑣) + 𝛾 ∗

𝑆𝑐𝑜𝑟𝑒𝐽𝑜𝑏𝐺𝑟𝑎𝑑𝑒(𝐽𝑣), where       1 3.1.3 Job Discovery

After measuring the job influence of each user in the social network, then the network of referral candidates will be expanded continuously from the job seeker. Specially, we denote as the social network which is constructed by nodes expanding for l layers,Θ𝑆𝑁 as the set of users of social network and the function Friends to express the friend set of certain user.

Following by this definition,

SN (0)=u, the job seeker, and Θ𝑆𝑁(0) = 𝑢 . 𝛩𝑆𝑁(1) = 𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢) and𝛩𝑆𝑁(2) = 𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)) too.

Along by the

SN (l)definition, we also use ECN (l) as the set of users included in the referral candidate network for l layers expanding, and Θ𝑅 as the set of referral candidates. The network expanding process in this stage can be described as:

(8) where Θ𝑅(0) = 𝜙 and R is the threshold of job influence level. In this research, we construct the network of referral candidate network after expanding three layers (l=3). To record the link from job seeker u to certain candidate v in the ECN, we will use the set

20

( , )

SocialPaths u v to record all the social paths from user u to user v:

1 2

( , ) { ( , ), ( , ), , _n( , )}

SocialPaths u v  SocialPath u v SocialPath u v SocialPath u v (9) For single path i, w denote as SocialPath u v and store all the nodes along this path. The _i( , ) notation will be



1 2



(u,v)= , , ,

i n

SocialPath sp sp sp where sp₁u and sp_n v. (10)

3.2 Social Appraisal Module

The objective in this stage is to evaluate and rank the best candidate as the reference of desired job. The model consists of two major components: willingness analysis and influence analysis. The former emphasizes how personal relationships affect the mission of reference; the latter stresses the relevance between the searching goal and the current resources, whether in personal side or the job side, which we have already computed before. We will discuss further about the details of each component in the following chapters and talk about the formula to measure each candidate in the end of 3.2.

3.2.1 Referral Willingness Analysis

We take a single link as the unit of referral willingness calculation. Among single link,

there are two components of referral willingness score: social tie strength and the similarity

of experience. We will mention how to accumulate the total willingness in the end of section.

3.2.1.1 Social Tie Strength Evaluation

In this study, the factors we considered about social tie strength includes, mutual friends,

interaction duration (day since the last communication), and status comment (the frequency

of like or comment in the status wall). The three factors represent three different dimension

21

discuss the details of each calculation in the following description.

From the structure view, the more mutual friends we have, the more possible there

would be a certain social link between us and then we are more likely to help each other.

From the intimacy view, since in this case we focus on the help of taking real referral action,

private messages exchanging will be a perfect index to reflect the actual interaction of

intimacy. Last, as for the intensity, we use the comment and like rate of status update to

consider how friends interact with each other.

Mutual Friends: The number of mutual friends between two people intuitively indicates

how close these people are in the view of social circle. Deduction from the concept of social

tie triangle, the stronger tie exists between two people, the more possible they would be

friends and get quite familiar with each other. Here we use the function MF(u,v) to record of

the number of mutual friends of user u and user v in order to compare the closeness between

these two people.

Interaction duration: According to the previous research, the social tie strength differs a lot

for the people talked once and for those who never talk [48]. By this finding, we record the

day since first communication and last communication of every pair as the index of duration. The function FirstDa u vy( , )is defined as the function which returns the time intervals from

( )

( )

( , )

,

( )

( )

Friends

u

Friends

v

u v

u

MF

Friends

Friends v







MF u( , )v 0 (11)

22

the first conversation to now. The functionLastDay u v means from the last conversation to ( , ) now. If the conversation just happened today, the value will be zero. If the conversation never

happens, both the function will return the infinite value.





1 1 ( , ) * 1 * , ( , ) ( , ) Duration FirstDay LastDay u v u v u v





   𝐹𝑖𝑟𝑠𝑡𝐷𝑎𝑦(𝑢, 𝑣) ≥ 0, 𝐿𝑎𝑠𝑡𝐷𝑎𝑦(𝑢, 𝑣) ≥ 0,

Status comment: There is another crucial element to affect the decision that whether we

would like to assist other or not. That is how we feel about others, the emotional aspect. If we

have good feelings about the other person, we are liable to give him or her hand. In

case of affection n dimension, we accumulated the number of like or comment a

user v have made on the status of person u during the two months to figure out how user v thinks about person u compared with his or her other friends. We useFriend_likeand

Friend_comment function to summarize how many times the user v has clicked like on the post of user u and how many times user v has left comment on the post of user u.

( )

( , )

( , ),

Comment Comment p Post u

Friend

u v

Post

v p







( )

( , )

( , ).

Like Like p Post u

Friend

u v

Post

v p







The set Post u( ) is used to record the all the status post of user u and function

PostLike(v, p)=1 if user v clicks like for the post p or PostLike(v, p)=0otherwise. The rule

works the same on Post_Comment function. If user v left comment on post p, then

PostComment(v, p)=1, otherwise PostComment(v, p)=0. We also conduct the linear normalization

(12)

(13)

(14)

(15) (16)

23

as below after calculating the score above:

𝐿𝑖𝑘𝑒_{𝑔𝑖𝑣𝑒𝑛}(𝑢, 𝑣) = 𝐹𝑟𝑖𝑒𝑛𝑑𝑙𝑖𝑘𝑒(𝑢,𝑣)−𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝑀𝑖𝑛 𝐹𝑟𝑖𝑒𝑛𝑑𝑙𝑖𝑘𝑒(𝑢,𝑓) 𝑀𝑎𝑥 𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝐹𝑟𝑖𝑒𝑛𝑑𝑙𝑖𝑘𝑒(𝑢,𝑓)−𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝑀𝑖𝑛 𝐹𝑟𝑖𝑒𝑛𝑑𝑙𝑖𝑘𝑒(𝑢,𝑓) 𝐶𝑜𝑚𝑚𝑒𝑛𝑡𝑔𝑖𝑣𝑒𝑛(𝑢, 𝑣) = 𝐹𝑟𝑖𝑒𝑛𝑑_{𝑐𝑜𝑚𝑚𝑒𝑛𝑡}(𝑢, 𝑣) − 𝑀𝑖𝑛 𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝐹𝑟𝑖𝑒𝑛𝑑𝑐𝑜𝑚𝑚𝑒𝑛𝑡(𝑢, 𝑓) 𝑀𝑎𝑥 𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝐹𝑟𝑖𝑒𝑛𝑑𝑐𝑜𝑚𝑚𝑒𝑛𝑡(𝑢, 𝑓) −𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝑀𝑖𝑛 𝐹𝑟𝑖𝑒𝑛𝑑𝑐𝑜𝑚𝑚𝑒𝑛𝑡(𝑢, 𝑓) ( , ) _given( , ) (1 ) * _given( , )

Status u v  



Like u v  



Comment u v . (17) Because the range of the value in _MF(u,v), _{Duration(u,v)} and _Status(u,v) does not

between zero to one, in order to decrease the error we apply the min-max normalization to

normalize those value in as shown in the equation(18), where value is the original value,

'

value is the new value after normalization, min is the minimum value of the population

and the max is the maximum value of the population.

min ' max min value value    . (18)

The social tie strength of user u and v is computed as:

𝑆𝑇(𝑢, 𝑣) = 𝛼 ∗ 𝑀𝐹(𝑢, 𝑣) + 𝛽 ∗ 𝐷𝑢𝑟𝑎𝑡𝑖𝑜𝑛(𝑢, 𝑣) + 𝛾 ∗ 𝑆𝑡𝑎𝑡𝑢𝑠(𝑢, 𝑣), 𝑤ℎ𝑒𝑟𝑒 𝛼 + 𝛽 + 𝛾 = 1

3.2.1.2 Experience Similarity Evaluation

Apart from the social tie strength, there is another similarity exerting in the real situation.

If the social tie means the private, personal and subjective side of willingness, the similarity

experience points the objective side. It is very common to see the job reference from the help

of alumni although the senior managers may not directly interact with the younger student. In

experience evaluation, two elements are considered: Living location and Education

24

background. The experience of user u and user v is computed as:

𝐸𝑆(𝑢, 𝑣) = 𝛼 ∗ 𝑆𝑖𝑚_{𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛}(𝑢, 𝑣) + (1 − 𝛼) ∗ 𝑆𝑖𝑚𝐿𝑜𝑐𝑎𝑡𝑖𝑜𝑛(𝑢, 𝑣) (20) The elements included in the evaluation are detailed as follows.

Living location: by observation and actual cases in the real world, it is possible for people to

take care of those who come from the same living areas in most referral cases. So in this case,

we take this factor into consideration.

1, if ( )= ( ) ( , ) 0 otherwise Location Location u Location v Sim u v    . (21) Education background: In the real world, it is common to find that the seniors who are

graduated are willing to share their own working experiences and even want to recruit some

talented juniors to join their companies. Proved by this fact, we are going to exam if the same

education background influences the effective of willing of referring. For the job reference

case, the collage, the graduated school and the major are the three information sources about

education experience we consider.

Sim_U(u,v)= 1, if University(u)=University(v)

0, otherwise ì í ï îï (22)

Sim_G(u,v)= 1, if Graduate(u)=Graduate(v)

0, otherwise ì í ï îï (23)

Sim_M(u,v)= 1, if Major(u)= Major(v)

0, otherwise ì í ï îï (24)

25

( , ) * ( , ) * ( , ) * ( , )

Education U G M

Sim u v  Sim u v  Sim u v  Sim u v (25) where     1.

After going through all the process of social tie strength evaluation and experience

similarity evaluation, now for a single link which connects user u and user v we can get the

score of willingness:

𝑊(𝑢, 𝑣) = 𝛼 ∗ 𝑆𝑇(𝑢, 𝑣) + (1 − 𝛼) ∗ 𝐸𝑆(𝑢, 𝑣) (26) The variable α and 1-α present the linear combination of the measurement.

3.2.1.3 Calculation of Referral willingness

After computing the willingness score among each social link, here we strive for finding

the willingness along the social path from job seeker u to certain candidate v. Combing the

result of each social link score and the social path we record in the 3.1.3, the calculation

process is depicted as:

( , ) ( , ) ( , ) ( , ) 1 ( , ) ( , ) 1 1 1

( , )

i i

SocialPath u v SocialPaths u v SocialPath u v SocialPaths u v n W sp sp W sp sp n i i i

W u v

Max

Max

    _ 





Besides locating the possible candidates in the second model of social search, here we

also strive for finding the best routing path, or the closest path to each possible candidate in

the goal of maximizing willingness in order to increase the possibility for those candidates to

help the job seekers.

The example of calculating process is described as follow Figure 6. After finding the

candidate Amy, the system has recorded that there are two social paths between from job

26

seeker John and Amy in the search stage. In this appraisal stage, the system will process

willingness score computing.

The willingness score between John and Amy will be computed as:

w (John, Amy)= w(John, Bob) * w (Bob, Andrew)* w(Andrew, Amy) =0.12*0.2*0.3=0.0072. w (John, Amy) = w(John, Doris) * w (Doris, Amy )=0.2*0.7=0.14

After processing this two social path from John to Amy, the system will automatically

update the highest score of social path, in this example, the path through Doris, for the

willingness of w (John, Amy).

3.2.2 Social Influence Analysis

In this stage, we mainly focus on analyzing the social influence. In the analysis of social

influence, we use three elements, total friends, social popularity and social engagement to

evaluate the status of certain user in the social circle. The first element is to measure the

possible social circle a person can impact on, and the second one is to measure the actual

attention a person can gain, or how popular he or she is in the social network. The last one is

to measure how this person becomes involved in the communities of social network. The Andrew 0.2 Bob 0.12 John Amy 0.3

Figure 6. The example of calculating willingness Doris

27

formula of social influence is described at the end of 3.2.2 and the details will be discusses

further in the following paragraph.

Total Friends: The concept of total friend is just to measure how many friends the user

has on the social network platform in order to know the scope he or she may influence and how many people that user can help distribute the job query. We use the tf u( ) function to

get the number of friends of certain user u. Considering the diversity of different social circle,

the normalization formula would be described as:

( ) ( ) ( ) ( ) ( ( )) ( ) ( ( )) ( ( )) v Friends u v Friends u v Friends u tf v Min tf v TF tf v tf v Max Min u       (28) Social Popularity: We expressed this kind of social power in the way of gaining

attention. Imagine that, maybe the person is not directly relevant with your desired job,

neither the company nor the industry. However, this popular person may publish your job

hunting news in his or her own social network and directs you to his or her friend as result,

which occurs a lot in the real world. In this evaluation, the average number of like and

comment from friends per status update are used to measure how much attention a user

would get through their personal social network. The definition is shown as:

𝐿𝑖𝑘𝑒_{𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑}(𝑢) = ∑ 𝑝∈𝑃𝑜𝑠𝑡(𝑢),𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝑃𝑜𝑠𝑡𝑙𝑖𝑘𝑒(𝑓,𝑝) 𝑛(𝑃𝑜𝑠𝑡(𝑢)) (29) 𝐶𝑜𝑚𝑚𝑒𝑛𝑡_{𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑}(𝑢) = ∑ 𝑝∈𝑃𝑜𝑠𝑡(𝑢),𝑓∈𝐹𝑟𝑖𝑒𝑛𝑑𝑠(𝑢)𝑃𝑜𝑠𝑡𝑐𝑜𝑚𝑚𝑒𝑛𝑡(𝑓,𝑝) 𝑛(𝑃𝑜𝑠𝑡(𝑢)) (30) The score of social popularity is generated as below:

28

Social Engagement: Here we talk about how a person participates the community in order to

know personal status in the social network. With the popularity of low-cost and

often-asynchronous social network on the Internet, social involvements turned from the local

and group-based to the internet-based. In this research, we consider the status in the way of participating group. We use the function group u( ) to denote the number of group user u

joins and the compare this involvement degree with user’s friends to conduct the

normalization from the view of social scope.

( ) ( ) ( ) ( ) ( ( )) ( ) ( ( )) ( ( )) v Friends u v Friends u v Friends u

group Min group v SE group v group v Max M u u in       (32) To sum up, the score of social influence will be defined as:

( ) * ( ) * ( ) * ( ), 1

SI u 



TF u 



SP u 



SE u where

  

   (33) Integrating the score of job influence in 3.1.2, here we could sum up and estimate the final

influence as the following formula:

𝐼(𝑢, 𝑣) = 𝛼 ∗ 𝐽𝐼(𝑢, 𝑣) + (1 − 𝛼) ∗ 𝑆𝐼(𝑢, 𝑣) (34) Note that because the social influence only considers personal power, so only the

candidate user v will be involved.

3.3 Weight Calculation

In this section we apply one of the most widely used approaches, the Analytic Network

29 Goal:

Criteria

Alternative

Step1: we design the hierarchy by identifying the goal, criterion, and alternatives like Figure

7. Our goal is to find the most proper referral for job seeker, and four elements in our model

would be our criterions, and the candidate set would be our alternatives. Level-1 weights are

the edges between criteria and goal, level-2 weights are the edges between alternatives and

criteria.

Step 2: In order to calculate the relative weight between each component, we collect the user

feedback and the criteria score from pervious criteria computing section as the data source,

and then we use pairwise comparison to compute the priority weight in level-1. First, we

form the pairwise comparison matrix shown at Formula 35:

𝑀_{𝑇𝐸𝐽𝑆} = [ 1 𝐸_𝑇𝐸 𝐸_𝑇𝐽 𝐸_𝑇𝑆 1 𝐸𝑇𝐸 1 𝐸𝐸𝐽 𝐸𝐸𝑠 1 𝐸𝑇𝐽 1 𝐸𝐸𝐽 1 𝐸𝐽𝑆 1 𝐸𝑇𝑆 1 𝐸𝐸𝑠 1 𝐸𝐽𝑆 1 ] (35)

where 𝐸𝑇𝐸 means relative weight ratio between social tie strength and experience

similarity, 𝐸𝑇𝐽 means relative weight ratio between social tie strength and job influence, 𝐸𝑇𝑆 Figure 7. ANP Structure

30

means relative weight ratio between social tie strength and social influence, 𝐸𝐸𝐽 means

relative weight ratio between experience similarity and job influence, 𝐸_𝐸𝑠 means relative

weight ratio between experience similarity and social influence, 𝐸𝐽𝑆 means relative weight

ratio between job influence and social influence. Then we use a mean method which shown at

Formula 36 to calculate the relative weight for each criteria:

4 4 1 1 1 4 ij i j ij i E W E   





(36)

where W means the relative weight of criteria i. By this formula then we can know how _i

these four components affect the final score to rank the proper candidate.

Back to the process of evaluating the proper candidate, after long processing each component,

we could finally generate the score of each candidate as well as the ranking list of users in the candidate set_R. The score of certain possible candidate v could be a proper reference for

user u to hunting a desire job is defined as follows:

_R(u,v)

=

W(u,v)* I(u,v)

(37)

W(u,v) stands for the willingness for user v to help user u, and I(u,v) means how user v could

influence the job user u wants as we mentioned previously. We will rank the candidate set by

this score and eventually output the job opening as well as a list of ranking candidate and the

social links to those people.

31

in Figure 8.

32

CHAPTER 4 EXPERIMENTS

After planning the whole framework of social referral mechanism, in this chapter, we conduct an experimental study and verify the effectiveness and efficiency of this proposed project. Generally speaking, the process of social referral is to request some information from social network, which expands by the connection of real social interactions, and the result is a list of people who may have the high possibility to help by the rank of recommendation calculation.

As for the input of this referral mechanism, since the experiment environment and the using habits of participants, we choose the most popular job website in Taiwan, 104 human resources bank and the most popular social services websites, Facebook, to execute the job-seeking process and observe the interaction process among people. The information of job openings is collected from the 104 human resources bank and the social interaction data is collected from Facebook. We use web crawler to analyze the job openings in the 104 websites and index some properties in advance in order to optimize the request process later happened in the social search among Facebook. As for each one who helps spread the job seeking request, we use the our own web-based app to gain more detailed personal information in order to index every user among the social circle of job seekers.

For technical side, we use Facebook API, such as the FQL and Graph API, to gather the real interactions among specific relationship to estimate the intimacy between friends or the popularity of certain people among his or her own social network. Furthermore, we implement our web-based app on the host of Bitbucket and platform of Heroku, which are both Git server and store our data in the Postgres database server. We utilize the analytical software, SPSS, as our tool to the factor analysis.

33

In the following sections, we will describe each procedure of data collection, storage and process as well as the calculation of how we actually use those collecting data to compute the recommendation results.

4.1 Experiment Process

The whole experiment is composed of four steps as Figure 4 illustration: network construction, social search for job query, social appraisal to rank candidates, output the final ranking list and request for feedback.

Step one: We will help the job seeker to distribute our web-based app to friends of his or

her social network on Facebook to construct the network. Besides providing personal information, we will also ask his or her friends, or called “participants” in our study, to help distribute the app to their friends as many as possible to maximize the social circle to collect more data. In this stage, considering the power of weak-tie in the issue of providing new job information as we mention in chapter 1, we could support this kind of distribution with a recommended list of people mixed of both strong-tie and weak-tie friends in their social network. Those two kinds of friends would be identified from the analysis of social tie strength in chapter 3.3.1. The interface of web-based app displays as Figure 8 and the recommended list is shown as Figure 9. Both personal information about each user and the information about each link are recorded in this stage.

Step two: The job seeker, in this case we call ”initiators” will choose preferred industry

or company as the keywords of the job request query and provide other personal information. A list of predefined industry, company and major categories will be provided to help the initiators to best describe the terms. The input interface shows as Figure 10.

34

Figure 16. The questionnaire distributed for participants

Figure 9. The suggestion list of distribution people Figure 8. The interface of web-based app for participants

35

Step three: After collecting the data, we start processing those data and calculating the

score of our four major modules explained in chapter 3. In this stage, we use the ANP

questions collected from the app to decide the weight of our formula and compute the score

of each component from the data we have collect.

Step four: Following calculating the score, we will demonstrate our result of ranking

list to job seeker, shown as Figure 11, and ask them to help evaluate the effective of our system by knowing how he or she think the candidates ranked by the score on the result list are willing to help him or her and in what percentage it is possible.

4.2 Data collection

We start our data collection by asking what kinds of jobs users prefer and index the job with predefined variable, such as industry, company, function. Here user can either enter the job query by his or her own preference or by the current available job openings listed on the job websites. Here we refer the categories of 104 job banking websites as our variable. Users also have to input their personal information, such as living location and education background Figure 10. The interface for job query

36

After figuring out the description about the ideal jobs, or we could call the “query condition” of our social search, then we collect the data by passing the webpage link we designed to gather all information we need in the further calculation. We use this web form to get both personal information concerning job and education and the records of social interaction happened in the Facebook. By the way, the automation of sending request can be done by the support of FQL.

There are 35 users as initiator involved, and for each of initiators we ask them to pick up three kinds of different types of jobs and spend approximately one week long to distribute and gather information from their social network. At the end of our experiment, we have 4,445 participants involved and collect 1,877,995 social links. On average, there are 127 participants of each the social network (the amount of people who fulfilled the web form from certain initiator) and 53657 social links among per social network. The details about the dataset summary show as follows:

Figure 11. The result of possible consulting candidate generates from the social referral system

37

Figure 11. Gender Distribution of Users

male 43% female

57%

Title Value

The Number of Initial Users 35 people

The Total Amount of Participants 4,445 participants

The Total Amount of Social Links 1,877,995 social links

The Number of Participants On Average Per User 127 participants

The Number of Social Links On Average

Per User

53657 social links

Table 1. The summary of dataset

4.2.1 User Profile

In order to make sure the sampling is fair and unified, we further analyze the

information about 35 users and 4,445 participants. Their gender distribution and age

distribution are shown respectively in Figure11, Figure 12, Figure 13 and Figure 14.

Figure 12. Gender Distribution of Participants

Figure 14. Age Distribution of Users

18~21 9% 22~28 41% 29~35 30% 36~40 13% 41up 7%

Figure 13. Age Distribution of Users

22~28 57% 29~35 29% 35~40 14%

38

Figure 15. Occupation distribution of participants

Figure 15. Occupation distribution of participants 28% 19% 3% 4% 5% 3% 1% 6% 9% 3% 6% 3% 2% 4% 2% 2% Informaiton Technology/Semiconduct Student Construction Retailers Government Entertainment/Publishing Transportation Finance Manufacters Accounting/Consulting Advertisement/Marketing Healthcare Education/Research Legal Tourism Property

Furthermore, we also do some statistics data about the occupation distribution of

participants and users ‘job queries as the following Figure 15 and Figure 16.

Figure 16. Occupation distribution of users ‘job queries

18% 18% 6% 6% 4% 17% 4% 11% 9% 7% Informaiton Technology/Semiconduct Student Construction Government Transportation Finance Manufacters Accounting/Consulting Advertisement/Marketing Healthcare

39 4.3 Weight Generation

In this research, we adapt ANP model to solve the weighting problem among all kinds of combination. Actually, we can tear down the whole model into 13 different variables in Figure 8. As we mentioned before, we use our app to know the relationship of weighting between the different factors from our participants. In order to ask more conveniently, we use the example questions like Table 2 to figure out the correlation between each pair of variables and convert it to the original calculation in Table 3.

Question I think when it comes to asking someone to help me find a job, with

the support from the one who is willing to help me has higher

possibility to successful referral than the one who has the influence to

help me. 5 4 3 2 1 Strongly Agree Strongly Disagree

After carefully analyzing, finally we can get the result from the collaborative opinions from all participants and thus use this to further recommend people of possible candidate lists with the method in chapter 3.4.

Table 3. ANP table for weighting Table 2. Question for estimating weighting

40

CHAPTER 5 RESULTS AND EVALUATION

In order to evaluate the accuracy of this proposed mechanism about recommendation referral candidate, we use the web-based app to keep tracking the following distribution process. After collecting data, we will use our system to execute the candidate processing and inform our users the candidate list. Furthermore, we ask them to rate the result of candidate and to see what’s their opinion about it as the feedback to enhance the mechanism.

5.1 Accuracy of Social Referral Information

In this experiment, top ten candidates are selected from the ANP result and ranked by

the score in the candidate list. The reason why we choose top ten people is because it’s the

reasonable number for job seeker to actually contact in person for specific job position. After

recommendation, we ask job seeker to review that list and pick up those people who job

seeker think are actually helpful as our evaluation. In this part of evaluation, we measure the

accuracy of the referral recommendation mechanism by the equation (38), where 𝛷_{𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠} is the set of referrals on the recommendation list and

𝛷_{𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠 ∩ ℎ𝑒𝑙𝑝𝑓𝑢𝑙 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠} is the set of referrals who job seeker thinks are truly useful in our recommended list.

𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦 =|𝛷𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠 ∩ ℎ𝑒𝑙𝑝𝑓𝑢𝑙 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠|

|𝛷𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑒𝑑 𝑟𝑒𝑓𝑒𝑟𝑟𝑎𝑙𝑠| (38)