Hybrid recommendations for mobile commerce
based on mobile phone features
Chuen-He Liou and Duen-Ren Liu
Institute of Information Management, National Chiao Tung University, Hsinchu, Taiwan Email: dliu@iim.nctu.edu.tw
Abstract: Mobile data communications have evolved as the number of third generation (3G) subscribers has increased. The evolution has triggered an increase in the use of mobile devices, such as mobile phones, to conduct mobile commerce and mobile shopping on the mobile web. There are fewer products to browse on the mobile web; hence, one-to-one marketing with product recommendations is important. Typical collaborative filtering (CF) recommendation systems make recommendations to potential customers based on the purchase behaviour of customers with similar preferences. However, this method may suffer from the so-calledsparsity problem, which means there may not be sufficient similar users because the user-item rating matrix is sparse. In mobile shopping environments, the features of users’ mobile phones provide different functionalities for using mobile services; thus, the features may be used to identify users with similar purchase behaviour. In this paper, we propose a mobile phone feature (MPF)-based hybrid method to resolve the sparsity issue of the typical CF method in mobile environments. We use the features of mobile phones to identify users’ characteristics and then cluster users into groups with similar interests. The hybrid method combines the MPF-based method and a preference-based method that uses association rule mining to extract recommendation rules from user groups and make recommendations. Our experiment results show that the proposed hybrid method performs better than other recommendation methods.
Keywords: mobile web, one-to-one marketing, product recommendation, collaborative filtering, mobile phone features, association rules
1. Introduction
In the last decade, mobile communications have evolved from 2G=2.5G to 3G=3.5G. As a result, the data transfer rate has been progressively upgraded from 64 Kbps (2.5G=GPRS) to 384 Kbps (3G=WCDMA), and 3.5 Mbps (3.5G= HSDPA), which is comparable to the wired Internet. The evolution has triggered an increase in the use of mobile devices, such as mobile phones, to conduct mobile commerce (m-commerce) on the mobile web (Chae & Kim, 2003; Venkatesh et al., 2003; Ngai & Guna-sekaran, 2007). M-commerce covers a large
num-ber of services, one of which is mobile shopping. Retailers have also increased their investment in mobile shopping channels to deliver dedicated products, content and promotions to customers.
Recommender systems have emerged in m-commerce or e-commerce applications to support product recommendation, which pro-vide individual product recommendation for each customer. Recommender systems assist business in implementing one-to-one marketing strategies, relying on customer purchase history to determine preferences and identify products that a customer may purchase. Recommender DOI: 10.1111/j.1468-0394.2010.00566.x
systems increase the probability of cross-selling, establish customer loyalty and fulfill customer needs by discovering products in which they may be interested in purchasing (Schafer et al., 2001). Recommender systems are widely used to recommend various items, such as consumer products, movies and music, to customers based on their interests (Hill et al., 1995; Shardanand & Maes, 1995). Generally, recommender sys-tems can be classified as collaborative or con-tent-based filtering techniques. Collaborative filtering (CF), which has been used successfully in various applications, utilizes preference rat-ings given by customers with similar interests to make recommendations to a target customer (Resnick et al., 1994; Linden et al., 2003; Lee, 2004; Cho et al., 2005; Liu & Shih, 2005). In contrast, content-based filtering derives recom-mendations by matching customer profiles with content features (Mooney & Roy, 2000; Martinez et al., 2007).
A number of product recommendation sys-tems have been developed for m-commerce on the mobile web (Kim et al., 2004; Choi et al., 2007; Lee & Park, 2007). For example, VISCOR is a mobile recommender that combines colla-borative and content-based filtering to provide better wallpaper recommendations (Kim et al., 2004). MCORE considers users’ context data to recommend mobile services (Choi et al., 2007). In addition, mobility information about user locations obtained from global positioning sys-tems (GPS) is usually used in m-commerce. A number of mobile recommendation systems use customers’ mobility patterns to make recommen-dations (Brunato & Battiti, 2003; Yang et al., 2008). Mobile phone features (MPF) such as Bluetooth and card slots have been used as product attributes to recommend mobile phone products. iTVMobi recommends mobile phone products based on the users’ preferences for MPF (Virvou & Savvopoulos, 2007). Existing works use MPF as product attributes to recom-mend mobile phone products, instead of using the features of mobile phones as the users’ characteristics (profiles) to recommend products. The typical CF method relies on finding users with similar interests to make recommendations.
However, it may suffer from the so-called spar-sity problembecause users only rate a few items. As a result, the user-item rating matrix is very sparse, so the recommendation quality is poor due to the difficulty of finding users with similar interests. In mobile shopping environments, active users may only browse=purchase a few items on the mobile web; thus, it is difficult to find users with similar interests based on the product preferences derived from users’ brow-sing=purchasing histories.
In this study, we propose a MPF-based hy-brid method to resolve the sparsity issue of the typical CF method used in mobile environ-ments. The MPF-based method uses the fea-tures of users’ mobile phones as user profiles to cluster users into groups with similar character-istics and then makes recommendations. The MPF indicate users’ motivations for using mo-bile services; thus, they can be used to identify users with similar product preferences. For example, the profiles of businessmen or sales representatives who own mobile phones with intelligence and GPS features may indicate a strong interest in high-tech 3C (Computer, Communication and Consumer) products. Thus, we consider MPF as user characteristics to help find users with similar interests. How-ever, some users who own mobile phones with the similar features may not have the similar product preferences. Hence, we still need to refer users’ product preferences for making recom-mendations. Thus, we propose a hybrid method which combines the MPF-based method and the preference-based method to improve recom-mendation quality by considering both MPF and product preferences. Similar to the MPF-based method, the preference-MPF-based method makes recommendations based on user groups that are clustered according to the users’ pro-duct preferences. Experiments were conpro-ducted to compare the performance of the proposed hybrid method with that of MPF-based, prefer-ence-based and typical CF methods. The results show that the hybrid method outperforms the other methods.
The remainder of this paper is organized as follows. In Section 2, we illustrate the
background of related methods. In Section 3, we describe the proposed MPF-based, preference-based and hybrid recommendation methods. In Section 4, we present the evaluation metrics and the experiment results. Then in Section 5, we summarize our findings and draw some conclusions.
2. Background
Our proposed method is based on MPF, and uses association rule-based and most frequent item-based recommendation methods. In this section, we briefly introduce the concepts and methods that are used in our research. This section also illustrates the typical CF method that is compared with our approach in experi-ment evaluation.
2.1. MPF
Mobile phones have evolved from the tradi-tional voice communication model to advanced digital convergence platforms with various fea-tures, such as Bluetooth technology, cameras, card slots, flash lights, as well as java, MP3, radio, touch screen, video and Wi-Fi functions (Ojanpera, 2006). These features enable users to access related mobile services, for example, download MP3 files, upload photos to blogs, video streaming and on-line shopping (Ko et al., 2007). Ling et al. (2006) investigated the impact of MPF on user satisfaction and analysed the feature preferences of diverse ethnic groups as well as preferences based on gender. Virvou and Savvopoulos (2007) developed an intelligent application called iTVMobi, which recommends mobile phone products on an interactive televi-sion. The system uses K-means clustering to group users based on their preferences for the attributes of mobile phones. The system then applies an association rule-based approach to recommend mobile phones based on the users’ preferences.
Existing works use MPF as product attributes to recommend mobile phone products, instead of using the features of mobile phones as user characteristics (profiles) to recommend
products. The features of different types of mobile phones can be obtained from the respec-tive websites. In this study, we log users’ mobile phone types when they browse products on a mobile shopping website. Then, we derive the phone features preferred by each user and use them to compile MPF-based user profiles. 2.2. Users clustering
Clustering techniques, which are usually used to segment users (Punj & Stewart, 1983; Chen et al., 1996), seek to maximize the variance among groups while minimizing the variance within groups. Many clustering algorithms have been developed, such as K-means, hierarchical and fuzzy c-means algorithms (Omran et al., 2007). K-means clustering (MacQueen, 1967) is a similarity grouping method widely used to partition a dataset into k groups. The K-means algorithm assigns instances to clusters based on the minimum distance principle, which assigns an instance to a cluster such that the distance to the centre of the cluster is the minimum over all kclusters.
2.3. Association rule-based recommendation method
Association rule mining tries to find the associa-tions between two sets of products in a transac-tion database. Agrawal et al. (1993) formalized the problem of finding association rules that satisfy the minimum support and the minimum confidence requirements. For example, assume that a set of purchase transactions includes a set of product items I. An association rule is an implication of the form: X) Y, where X I, Y I and XTY¼ F. X is the antecedent (body) and Y is the consequent (head) of the rule. Two measures, support and confidence, are used to indicate the quality of an association rule. The support of a rule is the percentage of transactions that contain both X and Y, whereas the confidence of a rule is the fraction of transactions that contain X and also contain Y.
Sarwar et al. (2000) described the association rule-based recommendation method as follows. For each customer, a customer transaction is
created to record all the products he=she pur-chased previously. An association rule mining algorithm is then applied to find all the recom-mendation rules that satisfy the given minimum support and minimum confidence. The top-N products to be recommended to a customer u, are then determined as follows: Let Xube the set
of products previously purchased by u. The method first finds all the recommendation rules X) Y in the rule set. If X Xu then all
products in Y Xu are deemed candidate
pro-ducts for recommendation to the customer u. The candidate products are then sorted and ranked according to the associated confidence of the recommendation rules, and the top-N candidate products are selected as the top-N recommended products.
2.4. Most frequent item-based recommendation method
The most frequent item-based recommendation method (Sarwar et al., 2000) counts the pur-chase frequency of each product by scanning the products purchased=browsed by the users in a cluster. Next, all the products are sorted by the purchase frequency in descending order. Final-ly, the method recommends the top-N products that have not been purchased=browsed by the target customer.
2.5. Typical CF method
CF (Resnick et al., 1994; Shardanand & Maes, 1995) utilizes the nearest-neighbour principle to recommend products to a target audience. The neighbours are identified by computing the similarity of customers’ purchase behaviour or tastes. The similarity is measured by Pearson’s correlation coefficient, which is defined as fol-lows: corrPðCi; CjÞ ¼ P s2IðrCi;s rCiÞðrCj;s rCjÞ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi P s2IðrCi;s rCiÞ 2P s2IðrCj;s rCjÞ 2 q ð1Þ whererCi andrCj denote the average number of
products purchased by customers Ci and Cj,
respectively; variable I denotes the mix of the
set of products; and rCi;s and rCj;s indicate,
respectively, that customers Ciand Cjpurchased
product item s.
The typical CF method utilizes k-nearest neighbours (k-NN) to recommend N products to a target user (Sarwar et al., 2000). The k-NN are identified by computing the similarity of customers’ purchase behaviour or tastes. The similarity is measured by Pearson’s coeffi-cient, as shown in equation (1). After the neighbourhood has been formed, the N recom-mended products are determined by the k-NN as follows. The frequency count of products is calculated by scanning the data about the products purchased=browsed by the k-NN. The products are then sorted based on the frequency count, and the N most frequent pro-ducts that have not been purchased by the target customers are selected as the top-N recommen-dations.
3. Proposed MPF-based hybrid recommendation method
In this section, we describe the proposed hybrid recommendation method, which combines an MPF-based method and a preference-based method, as shown in Figure 1. First, the MPF-based method extracts the features of users’ mobile phones from the respective phone web-sites, as shown on the left-hand side of the figure. The features of users’ mobile phones are taken as user profiles to identify users with similar characteristics. The system then applies the K-means clustering method to cluster users into groups based on the similarity of users’ MPF. Next, the association rules and frequently browsed products are extracted from each clus-ter. The system then recommends products based on the association rules and frequently browsed products. However, there may be very few products recommended according to the association rules because of the limited number of products that can be browsed on the mobile web. If the association rule-based recommenda-tions are not sufficient, the most frequent item-based recommendations are used to recommend
products to users. Similar to the MPF-based method, the preference-based method, shown on the right-hand side of Figure 1, clusters users by the K-means clustering method based on Pearson’s correlation coefficient of users’ pro-duct preferences. It then recommends propro-ducts based on the association rules and the most frequent items. Finally, the hybrid recom-mendation scheme combines the MPF-based recommendations and preference-based recom-mendations with the hybrid ratio determined by the preliminary analytical data to recommend products. We discuss the recommendation phase of the MPF-based, preference-based and hybrid recommendation schemes in Sections 3.2 and 3.3, respectively.
3.1. Data pre-processing and clustering
We obtained the features of each mobile phone from one of the mobile phone websites. There are more than 100 features on a mobile phone. It is hard to analyse all of them. Therefore, we selected the features based on the following three criteria. (1) Advertisements of a mobile phone retailer: The advertisements of a mobile phone retailer often list the important features for users’ preferences and comparison; (2) Fea-tures with too many missing values are not suitable for analysis and thus are not selected; and (3) Features with values that can discrimi-nate the differences of mobile phones. Table 1 lists the selected features, including Bluetooth Figure 1: An overview of the proposed hybrid recommendation scheme.
technology, cameras, card slots, flash lights, as well as java, MP3, radio and video functions.
The price feature is complicated for analysis, since the prices of mobile phones may vary under different subscription fees provided by various service providers. Thus, we do not select the price feature. The price feature has been somewhat implicitly considered and depends on the selected eight features, because mobile phones with more features are often more ex-pensive. The display feature is not listed in the advertisements of the mobile phone retailer and is a combination of three discrete data type features including screen size, colour and mate-rial. These values of the display features are missing and are difficult to collect. Thus, we do not select the display feature.
We calculate the similarity of users based on the selected features. The camera quality feature, which is a discrete data type, and the other seven features are Boolean data types. The camera resolution pixels (3.2, 2.0 and 1.3 mega-pixel resolution) need to be normalized to the seman-tic values of high, medium and low, as shown in equation (2) (Lin et al., 2003). Therefore, we use three Boolean operators to represent high, med-ium and low quality camera resolution (1, 0, 0) represents high quality (0, 1, 0) represents med-ium quality, and (0, 0, 1) represents low quality.
Zcamera¼
Xcamera MðXcameraÞ
sXcamera ð2Þ
where Xcamera is the camera quality; and
M(Xcamera) and sXcamera are, respectively, the
mean value and the standard deviation of the camera quality.
Next, we identify all the users’ mobile phones and expand the phones’ features to form a user-mobile phone feature matrix, as shown in Table 2. In the matrix, the values of the camera resolution mega-pixels are transformed into semantic values based on equation (2), with Zcamera< 0.8, 0.8 % Zcamera% 0.8 and
Zcamera> 0.8, representing low-level,
medium-level and high-medium-level quality cameras, respec-tively. We then use the matrix to cluster the users into groups. The MPF-based method clusters users by the K-means clustering method with Pearson’s correlation coefficient based on the users’ preferred MPF.
User product preference clustering is more intuitive than user MPF clustering, as it clusters users directly based on the user-product prefer-ence matrix. The preferprefer-ence-based method clus-ters users by the K-means clustering method with Pearson’s correlation coefficient based on users’ product preferences.
3.2. The MPF-based and preference-based recommendation phase
After clustering users into groups based on similar MPF or product preferences, the asso-ciation rules and the most frequent items in each group (cluster) are generated for the next step of the recommendation phase. The steps of the MPF-based and preference-based recommenda-tion phase are shown in Figure 2 and described as follows. Let Xurepresent the set of products
browsed previously by a user u. For each asso-ciation rule Xk! Yk, if Xk Xu then all
pro-ducts in Yk Xu, denoted by Yuk, are regarded as
candidate products for recommendation to the user u. Let YuAR be the set of all candidate
products generated from all association rules that satisfy Xk Xu. The products in YuAR are
ranked according to c(Yuk), that is, the associated
confidence of the association rule (AR) Xk! Yk.
We compare the number of candidate pro-ductsjYARu jand the top-N recommendations. If the former is greater than the latter, the system Table 1: Mobile phone features
No Feature Data type Value
0 Bluetooth Boolean (0, 1)
1 Camera
quality
Discrete (Low, Medium, High)
2 Card slot Boolean (0, 1)
3 Flash light Boolean (0, 1)
4 Java Boolean (0, 1)
5 MP3 Boolean (0, 1)
6 Radio Boolean (0, 1)
Table 2: User-mobile phone feature matrix
User ID Phone type Bluetooth
Camera
Card slot Flash light Java MP3 Radio Video H M L
1 MOTO V191 0 0 0 1 0 0 1 1 0 1
2 Nokia N70 1 1 0 0 1 1 1 1 1 1
3 SAMSUNG SGH-Z238 1 0 1 0 1 0 1 1 0 1
4 Sony Ericsson K800i 1 1 0 0 1 1 1 1 1 1
recommends the top-N products among the products in YuAR. On the other hand, if the
number of candidate productsjYARu j is less than the number of top N recommendations ðjYAR
u j < NÞ, the remaining N jYARu j
pro-ducts for recommendation are selected from YuMF. The selected products are the most
fre-quent items ranked according to the frequency count of products browsed by the users in the target user’s cluster. Then, products in YuMFthat
have not been browsed by the user u and have not been included in YuAR are added to the
recommended product list so that the number of top-N recommendations is sufficient.
3.3. The hybrid recommendation phase
The hybrid recommendation phase combines the MPF-based method and the preference-based method, as shown in Figure 3. Similar to the MPF-based method, the hybrid method first recommends products based on the association rules (AR); and then recommends products based on the most frequent item (MF) count. Let XMi! YMiand XPj! YPj be the associa-tion rules extracted from an MPF-based cluster (M) and a preference-based cluster (P), respec-tively; and let their associated confidence scores be cMiand cPj, respectively. In addition, let Xu
represent the set of products previously browsed by the target user u; and let YuARbe the set of all
candidate products generated from all associa-tion rules that satisfy XMi Xu or XPj Xu.
The products in YuAR are ranked according to
the weighted sum of their confidence scores. cH¼ wM cMiþ wP cPj ð3Þ
where wMand wPare the weights assigned to the
MPF-based approach and the preference-based approach, respectively.
Similar to the MPF-based method and the preference-based method, if the number of can-didate productsjYARu jis less than the number of top N recommendations ðjYAR
u j < NÞ, the
re-maining N jYAR
u j products for
recommenda-tion are selected from YuMF. The selected
products are the most frequent items, which are ranked according to the frequency count of
products browsed by the users in the target user’s MPF cluster and preference cluster. The most frequent items are ranked as follows. Let YMF M and YMF P denote the set of most frequent items derived from the target user’s MPF-cluster and preference cluster, respec-tively; and let fcM and fcP represent the fre-quency count of an item in YMF M and YMF P, respectively. Products in YuMFthat have
not been browsed by the target user u and have not been included in YuAR are recommended
based on the ranking order of the weighted sum of their frequency counts.
fcH¼ wM fcMþwP fcP ð4Þ
where wMand wPare the weights assigned to the
MPF-based approach and the preference-based approach, respectively.
The relative effects of the MPF-based ap-proach and preference-based apap-proach on the recommendation quality may be different for different top-N recommendations; therefore, we set different values of wMand wP. We discuss the
effects in detail in the next section.
4. Experiment evaluation
4.1. Experiment setup and data sets
Data for the mobile web log was collected between October 2006 and January 2007. The dataset, which contained information about 1692 users, 1416 products and 184 mobile phones, was divided as follows: 80% for training and 20% for testing. The training set was also used as the data set in the preliminary analytical experiment. Specifically, 55% of the data set was used to derive recommendation rules; and 25% was used as a preliminary analytical data set to determine the number of clusters, the feature combinations, and the hybrid weights assigned to the MPF-based and preference-based meth-ods based on the quality of the recommenda-tions. There were 1353 users and 165 mobile phones in the training data set, and 339 users and 93 mobile phones in the test data set. The
minimum support was set at 0.004, and the minimum confidence level was set at 0.6.
4.2. Evaluation metrics
Two metrics, precision and recall, are com-monly used to measure the quality of a
recom-mendation. They are also used in the field of information retrieval (Van Rijsbergen, 1979; Salton & McGill, 1986). Product items can be classified into products that customers are inter-ested in browsing, and those that they are not interested in browsing. The recommendation method then recommends products to the Figure 3: The hybrid recommendation phase.
customers accordingly. The recall metric indi-cates the effectiveness of a method for locating interesting products, while the precision metric represents customers’ level of interest in the recommended product items.
Recall is the fraction of interesting product items that can be located.
Recall¼number of correctly recommended items number of interesting items
ð5Þ Precision is the fraction of the recommended products that customers find interesting.
Precision¼number of correctly recommended items number of recommended items
ð6Þ The items that were considered interesting to customers were the products the customers browsed in the test set. Correctly recommended items were those that matched interesting items. Because increasing the number of recommended items tends to reduce the precision and increase the recall, the F1 metric is used to balance the tradeoff between precision and recall (Van Rijs-bergen, 1979). The F1 metric, which assigns equal weights to precision and recall, is given by
F1¼2 recall precision
recallþ precision ð7Þ Each metric was computed for each user, and the average value was computed for each clus-ter. The overall average (i.e. of all users) was calculated to measure the quality of the recom-mendations.
4.3. Experiment results
4.3.1. MPF and cluster number selection Al-though we selected eight MPF in Section 3.1, the recommendation quality may not be the best if we combine all the features. Therefore, we try all possible combinations of the eight features to determine the best combination and the number of clusters. We cluster users by the K-means clustering method with Pearson’s correlation coef-ficient based on the selected features. Using the MPF-based method described in Section 3.2, we
try various MPF combinations and various num-bers of clusters between two and eight. The best recommendation quality of the preliminary analy-tical data is derived by combining five features, namely the Bluetooth, card slot, flash light, java and video functions, and the number of clusters is three. Hence, we use these five features with three clusters as the parameters for MPF-based recom-mendation.
4.3.2. Mobile phone and product preference clus-ter identification Based on the results derived in the previous section, we divide the training users into three clusters according to the five selected phone features, as shown in Table 3. The Bluetooth function enables users to connect to the other Bluetooth devices, including ear-phones and notebooks. The card slot function expands a mobile phone’s data storage capacity for music, photos and movie files. The flash light function improves the quality of photographs taken in certain environments. Mobile phones with the java function can run java applications, including games; while phones with a video function are becoming increasingly popular for playing MP4 and movies in 3GP format.
Based on Table 3, we can calculate the feature frequency of each cluster by considering the frequency count and the representative MPF of each cluster that are above the frequency thresh-old of 50%. The frequency count of a feature is defined as the number of users’ phones that have the feature divided by the number of users. According to the feature frequency of each cluster, the users are classified into three types of MPF. Users in cluster 0 prefer camera phones with advanced features, such as Bluetooth, card slot and flash light functions; users in cluster 1 prefer simple phones with basic java features and video functions; and users in cluster 2 prefer feature phones with Bluetooth and card slot functions for device connectivity and data storage.
The preference-based method clusters users according to their product preferences, that is, products browsed by users. We cluster users into four groups based on the best recommendation
quality achieved using the preliminary analyti-cal data set. We use the product category fre-quency count with at threshold of 20% to identify the characteristics of each product pre-ference cluster, as shown in Table 4. The fre-quency count of a product category is defined as the number of users that browse the product category divided by the total number of users.
4.3.3. Determining the MPF and preference weights of the hybrid recommendation sche-me The hybrid recommendation scheme is based on the hybrid weighting ratios wM and
wP (wP¼ 1 wM) of the mobile phone and
product preference clusters. Hybrid recommen-dation becomes pure preference-based recom-mendation when wM equals zero, and pure
MPF-based recommendation when wM equals
one.
The top-N recommendations are divided into two segments. One segment is from the top-1 to the top-10 recommendations and the other is from the top-11 to the top-20 recommendations. We choose the top-5 and top-15 recommenda-tions to represent the first and second segments, respectively. The quality of the top-5 and top-15 hybrid recommendations with different MPF weights (wM) is shown in Figure 4. The best
recommendation quality for the 5 and top-15 occurs when wM¼ 0.9 and wM¼ 0.6,
respec-tively. We use these weights as the hybrid weighting ratios of the hybrid recommendation scheme in Section 4.3.4.
4.3.4. Evaluation of MPF-preference hybrid re-commendation methods We compare two pro-posed methods, namely, MPF-based and Hybrid MPF-Preference methods, with the other two methods, preference-based and CF methods. MPF-based method cluster users into groups based on users’ MPF and recommend products according to the association rules and most frequent items extracted from user groups. Preference-based method makes recommenda-tions based on user groups that are clustered according to the users’ product preferences.
Table 3: Mobile phone cluster classification Cluster ID Users Bl uetooth Frequenc y (%) Card slo t Fre quency (%) Flash light Frequency (%) Jav a Frequency (%) Vi deo Frequenc y (%) Mobile phone features Phone type 0 449 442 98 368 82 400 89 442 98 439 98 Java, Video, Bluetooth, Card slot, Flash light Camera phone 1 319 0 0 7 2 13 4 318 100 301 94 Java, Video Simple phone 2 162 156 96 162 100 8 5 156 96 146 90 Java, Video, Bluetooth, Card slot Feature phone Total 930 598 64 537 58 421 45 916 98 886 95
Hybrid MPF-Preference recommendations are generated by a combination of the MPF-based and preference-based recommendation schemes with the hybrid weighting ratio described in Section 3.3. The hybrid weighting ratio de-scribed in Section 4.3.3 is set at wM¼ 0.9 for the
first top-N segment (top1-10) and wM¼ 0.6 for
the second top-N segment (top11–20). The CF method is a typical k-NN CF method that recommends the top-N most frequently occur-ring products of the k-NN (similar users). Be-cause the average number of users in the product clusters is 232.5(¼ 930=4), we choose k¼ 200 as the number of nearest neighbours. Table 5 presents the precision, recall and F1-metric evaluation of k-NN CF, Preference-based, MPF-based and Hybrid MPF-Preference methods.
The F1 values of all methods are low, since the user-item matrix of our experiment data is very sparse. Although the F1 values of our proposed methods are still low, our methods
can achieve better improvement over conven-tional methods. For example, as listed in Table 5, the average F1-metric of the MPF-based method is 11% better than the prefer-ence-based method. Furthermore, the average F1-metric of the hybrid MPF-Preference meth-od, which combined MPF-based and prefer-ence-based methods, is 33% better than the preference-based method. The F1-metric of the hybrid MPF-Preference, MPF-based, Prefer-ence-based and k-NN CF methods are shown in Figure 5.
As shown in Figure 5, the recommendation quality of all the methods declines after the top-4 recommendations, as the number of re-commended products increases. Recall that as-sociation rule-based recommendations are based on the items users browsed previously. There are only a few recommended products because the average number of previously browsed products was 3.87. Therefore, the most frequent item recommendations are used to support the association rule recommendations if the number of recommended products is not sufficient. However, most frequent item-based recommendations are not better than associa-tion rule-based recommendaassocia-tions, so the recom-mendation quality deteriorates after the top-4 recommendations.
4.3.5. The effect of the hybrid method on mobile phone and product preference clusters Figure 6 shows that the mobile phone cluster 0 (camera phones with java, video, Bluetooth, card slot and flash light functions) achieves the best recommendation quality, followed by cluster 1 (simple phones with java and video functions), Table 4: Product preference cluster identification
Cluster ID Users Product category
0 185 Lingerie, pants
1 336 Mobile phones, cordless phones, digital cameras
2 179 Hotels, travel coupons, food, domestic travel
3 230 Skin care, MP3, cosmetics, living products
Total 930 Hybrid recommendations 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 wM F1-metric top-5 top-15
Figure 4: The weighting ratiowMof the hybrid
and cluster 2 (feature phones with java, video, Bluetooth and card slot functions). Among all the phone types, camera phones with the flash light feature yield the best recommendation quality. The owners of camera phones like to browse for digital cameras and travel products because they like to travel and take photo-graphs. We also evaluate the effect of the hybrid method on the recommendation quality of MPF-based clusters. Figure 6 shows that the recommendation quality of the hybrid method (hybrid 0–2) is better than that of the based method (mphone 0–2) for each MPF-based cluster. In other words, the effect of
combining MPF-based recommendations with preference-based recommendations is positive.
From Figure 7, we observe that product cluster 0 (lingerie, pants and skincare products) achieves the best recommendation quality in terms of product preferences, followed by pro-duct cluster 2 (hotels, travel coupons, food and domestic travel), product cluster 1 (mobile phones, digital cameras, cordless phones and notebooks), and product cluster 3 (skincare, mp3, cosmetics and consumer products). Users who prefer lingerie and underwear products receive better quality recommendations than users who prefer other products. We also Table 5: Evaluation of k-NN CF, Preference-based, MPF-based and hybrid methods
TopN k-NN CF Preference-based MPF-based Hybrid MPF-preference
Precision Recall F1 Precision Recall F1 Precision Recall F1 Precision Recall F1 2 0.015 0.004 0.006 0.153 0.085 0.092 0.161 0.088 0.099 0.176 0.100 0.110 4 0.026 0.017 0.017 0.104 0.113 0.089 0.122 0.128 0.106 0.140 0.157 0.125 6 0.036 0.055 0.035 0.080 0.124 0.080 0.096 0.156 0.100 0.113 0.186 0.118 8 0.039 0.098 0.045 0.072 0.146 0.081 0.079 0.165 0.091 0.092 0.195 0.106 10 0.035 0.107 0.044 0.063 0.156 0.076 0.067 0.172 0.082 0.081 0.212 0.100 12 0.030 0.109 0.040 0.057 0.165 0.072 0.058 0.178 0.075 0.072 0.221 0.094 14 0.027 0.111 0.036 0.051 0.171 0.066 0.051 0.180 0.069 0.064 0.227 0.087 16 0.023 0.112 0.033 0.046 0.174 0.062 0.045 0.181 0.063 0.058 0.236 0.081 18 0.021 0.112 0.030 0.042 0.179 0.059 0.043 0.197 0.062 0.053 0.244 0.077 20 0.019 0.112 0.028 0.040 0.187 0.057 0.041 0.210 0.061 0.049 0.248 0.073 Average 0.027 0.084 0.032 0.071 0.150 0.073 0.076 0.165 0.081 0.090 0.203 0.097 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Top-N F1-metric Hybrid MPF-Preference MPF-based Preference-based k-NN CF
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Top-N F1-metric hybrid0 mphone0 hybrid1 mphone1 hybrid2 mphone2
Figure 6: Effect of the hybrid method on the recommendation quality for mobile phone clusters.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Top-N F1-metric hybrid0 product0 hybrid2 product2 hybrid1 product1 hybrid3 product3
evaluate the effect of the hybrid method on the recommendation quality of preference-based clusters. Figure 7 shows that the recommenda-tion quality of the hybrid method (hybrid 0–3) is better than that of the preference-based method (product 0–3) for each preference-based cluster. Hence, combining the preference-based method with the MPF-based method can improve the quality of recommendations.
5. Conclusion
We have proposed a MPF-based hybrid method to resolve the sparsity issue of the typical CF method in mobile environments. We assume that the MPF preferred by users indicate their interest in particular m-commerce products and services; thus, they can be used to group users with similar interests. The hybrid method com-bines the MPF-based method and preference-based method, which uses association rule mining to extract recommendation rules from user groups and make recommendations.
Experiment results show that the quality of MPF-based recommendations is better than that of the preference-based method and the typical k-NN CF scheme. However, the hybrid method outperforms the MPF-based, prefer-ence-based and the typical k-NN CF methods.
According to the cluster analysis results, mobile phone cluster 0 (camera phones with Bluetooth, card slot, flash light, java and video functions) yields the best recommendation quality among the mobile phone clusters; product cluster 0 (lingerie, pants and skincare products) achieves the best recommendation quality in terms of product preference clusters. The hybrid method, which combines recommendations derived by the MPF-based and preference-based methods, im-proves the recommendation quality of MPF-based clusters and preference-MPF-based clusters.
Acknowledgements
This research was supported in part by the National Science Council of the Taiwan under grant NSC 96-2416-H-009-007-MY3.
References
AGRAWAL, R., T. IMIELIN´SKI and A. SWAMI (1993) Mining Association Rules between Sets of Items in Large Databases, Washington, DC, USA: ACM SIGMOD International Conference on Manage-ment of Data, 207–216.
BRUNATO, M. and R. BATTITI(2003) PILGRIM: A location broker and mobility-aware recommenda-tion system. Proceedings of the First IEEE Interna-tional Conference on Pervasive Computing and Communications (PerCom 2003). Fort Worth, TX, USA, pp. 265–272.
CHAE, M. and J. KIM(2003) What’s so different about
the mobile Internet?, Communications of the ACM, 46, 240–247.
CHEN, M.S., J. HANand P.S. YU(1996) Data Mining:
an overview from a database perspective, IEEE Transactions on Knowledge and Data Engineering, 8, 866–883.
CHO, Y.B., Y.H. CHO and S.H. KIM(2005) Mining
changes in customer buying behavior for collabora-tive recommendations, Expert Systems with Applica-tions, 28, 359–369.
CHOI, J.Y., H.S. SONGand S.H. KIM(2007) MCORE:
a context-sensitive recommendation system for the mobile Web, Expert Systems, 24, 32–46.
HILL, W., L. STEAD, M. ROSENSTEINand G. FURNAS
(1995) Recommending and Evaluating Choices in a Virtual Community of use. Proceedings of the SIG-CHI Conference on Human Factors in Computing
Systems, Denver, CO, USA: ACM Press
=Addison-Wesley Publishing, 194–201.
KIM, C.Y., J.K. LEE, Y.H. CHO and D.H. KIM
(2004) Viscors: a visual-content recommender
for the mobile web, IEEE Intelligent Systems, 19, 32–39.
KO, S.M., Y.G. JIand D. KIM(2007) Understanding
influence of mobile internet services on life behavior of mobile users, Lecture Notes in Computer Science, 4553, 944–953.
LEE, H.J. and S.J. PARK (2007) MONERS: a news
recommender for the mobile web, Expert Systems with Applications, 32, 143–150.
LEE, W.-P. (2004) Applying domain knowledge and
social information to product analysis and recom-mendations: an agent-based decision support sys-tem, Expert Systems, 21, 138–148.
LIN, Q., Y. CHEN, J. CHENand Y. CHEN(2003) Mining
inter-organizational retailing knowledge for an alli-ance formed by competitive firms, Information and Management, 40, 431–442.
LINDEN, G., B. SMITHand J. YORK(2003) Amazon. com recommendations: item-to-item collaborative filtering, IEEE Internet computing, 7, 76–80.
LING, C., W. HWANGand G. SALVENDY(2006)
features, Universal Access in the Information Society, 5, 239–249.
LIU, D.-R. and Y.-Y. SHIH (2005) Integrating AHP
and data mining for product recommendation based on customer lifetime Value, Information & Manage-ment, 42, 387–400.
MACQUEEN, J. (1967) Some Methods for Classification
and Analysis of Multivariate Observations. Fifth Ber-keley Symposium on Mathematical Statistics and Probability, Berkeley, CA: Statistical Laboratory of the University of California, Berkeley, University of California Press, 281–297.
MARTINEZ, L., L.G. PEREZand M. BARRANCO(2007) A multigranular linguistic content-based recommen-dation model, International Journal of Intelligent Systems, 22, 419–434.
MOONEY, R.J. and L. ROY(2000) Content-based book recommending using learning for text categorization, in Proceedings of the Fifth ACM Conference on Digital Libraries, San Antonio, TX, USA: ACM, 195–204.
NGAI, E.W.T. and A. GUNASEKARAN(2007) A review
for mobile commerce research and applications, Decision Support Systems, 43, 3–15.
OJANPERA, T. (2006) Convergence transforms internet, Wireless Personal Communications, 37, 167–185.
OMRAN, M.G.H., A.P. ENGELBRECHTand A. SALMAN
(2007) An overview of clustering methods, Intelligent Data Analysis, 11, 583–605.
PUNJ, G. and D.W. STEWART(1983) Cluster analysis
in marketing research: review and suggestions for application, Journal of Marketing Research, 20, 134–148.
RESNICK, P., N. IACOVOU, M. SUCHAK, P. B ERG-STROM and J. RIEDL (1994) GroupLens: an open architecture for collaborative filtering of netnews, in Proceedings of the 1994 ACM Conference on Compu-ter Supported Cooperative Work, Chapel Hill, NC, USA: ACM, 175–186.
SALTON, G. and M.J. MCGILL(1986) Introduction to Modern Information Retrieval, NY, USA: McGraw-Hill.
SARWAR, B., G. KARYPIS, J. KONSTANand J. RIEDL
(2000) Analysis of recommendation algorithms for e-commerce, in Proceedings of the Second ACM Conference on Electronic Commerce, Minneapolis, MN, USA: ACM, 158–167.
SCHAFER, J., J. KONSTAN and J. RIEDL (2001)
E-commerce recommendation applications, Data
Mining and Knowledge Discovery, 5, 115–153. SHARDANAND, U. and P. MAES (1995) Social
information filtering: algorithms for automating ‘‘word of mouth’’, in Proceedings of the SIGCHI
conference on Human factors in computing systems,
Denver, CO, USA: ACM=Addison-Wesley
Publish-ing, 210–217.
VAN RIJSBERGEN, C.J. (1979) Information Retrieval, MA, USA: Butterworth-Heinemann Newton. VENKATESH, V., V. RAMESHand A.P. MASSEY(2003)
Understanding usability in mobile commerce, Com-munications of the ACM, 46, 53–56.
VIRVOU, M. and A. SAVVOPOULOS(2007) An intelli-gent TV-shopping application that provides recom-mendations. ICTAI 2007, 19th IEEE International Conference on Tools with Artificial Intelligence, Patras, Greece, pp. 366–373.
YANG, W.S., H.C. CHENG and J.B. DIA (2008) A
location-aware recommender system for mobile shopping environments, Expert Systems with Appli-cations, 34, 437–445.
The authors
Chuen-He Liou
Chuen-He Liou is a PhD student of the Institute of Information Management at the National Chiao Tung University of Taiwan. He received the MS degree in Physics from the National Taiwan Normal University. His research inter-ests include information systems, recommender systems, mobile commerce, channel marketing and strategic planning.
Duen-Ren Liu
Dr Duen-Ren Liu is a Professor of the Institute of Information Management at the National Chiao Tung University of Taiwan. He received the BS and MS degrees in Computer Science from the National Taiwan University and his PhD in Computer Science from the University of Minnesota. His research interests include information systems, knowledge engineering and management, workflow systems, electronic commerce and recommender systems.
