CHAPTER 3 Proposed Approach
3.4 An Example
In this section, we use an example to illustrate the proposed algorithm to mine a diverse group stock portfolio with investor sentiment index. Assume there are fifteen companies. The trading data of one of the fifteen companies s1 is shown in Table 2. For the convenience of the example, we define the Low BSI threshold as PR20(BSIt) and the High BSI threshold as PR80(BSIt). Actually, the Low BSI threshold is PR10(BSIt) and the High BSI threshold is PR90(BSIt) in the experiment. Then we follow these steps to proceed with the algorithm.
STEP 1: Generate the trading signal on stock using the following sub-steps:
Sub-step 1.1: Calculate the BSIt on every trading day using the formula (1) to (3). Assume that there is a company trading data, and the related data is shown in Table 7.
Table 7. Trading data on the stock from 2011/1/3 to 2011/1/14.
Trading day Stock price BML BSL V BI SI DT
2011/1/3 12 62 59 456 263 196 6.50%
2011/1/4 17 28 79 801 217 356 8.60%
2011/1/5 10 65 66 865 151 13 8.00%
2011/1/6 15 26 60 807 390 400 7.30%
2011/1/7 14 66 57 451 196 155 5.40%
2011/1/10 13 35 72 738 205 309 7.80%
2011/1/11 14 12 20 564 133 243 7.80%
2011/1/12 16 33 57 798 201 260 7.30%
2011/1/13 13 98 46 398 157 133 6.60%
2011/1/14 20 68 23 766 135 192 5.90%
On 2011/1/3, the VBt is 193 (= 456-263), the VSt is 260 (= 456-196) and the BSIt is 0.006623 (= (62-59)/(193+260)). In the same way, we get the all BSIt on every trading day.
Sub-step 1.2: Calculate the high BSI threshold and low BSI threshold using the following
sub-steps:
Sub-step 1.2.1: Rank all BSIt and get the PR value of all BSIt by the formula (5).
Sub-step 1.2.2: Get the PR80 of BSIt to be the high BSI threshold and get the PR20 of BSIt
to be the low BSI threshold.
Sub-step 1.2: Calculate the difference between DTt-1 and DTt on every trading day.
Sub-step 1.3: Generate the original trading signal on every trading days by using Rule1 and Rule2. The result is shown in Table 8.
Table 8. The investor sentiment indices from 2011/1/3 to 2011/1/14.
Trading day DTchange BSI RankBSI BSI-PRvalue Trading point
2011/1/3 0.006623 8 20
2011/1/4 2.10% -0.04956 1 90 Buy
2011/1/5 -0.60% -0.00064 7 30
2011/1/6 -0.70% -0.04126 2 80
2011/1/7 -1.90% 0.016334 6 40
2011/1/10 2.40% -0.03846 5 50
2011/1/11 0.00% -0.01064 4 60
2011/1/12 -0.50% -0.02115 3 70
2011/1/13 -0.70% 0.102767 9 10 Sell
2011/1/14 -0.70% 0.037344 10 0 Sell
STEP 2: Get the information about Buy price as 17 and the Sell price as 20 on this stock s1. Back to STEP 1 to get the buying price and selling price until all the buying price and selling price is obtained on fifteen companies.
With the buying prices and selling prices on fifteen companies, assume related data are shown in Table 9.
Table 9. The data of stocks used in this example.
Stock Buy price Sell price Cash dividends Type varCD Capital
s1 17 20 0.46 0 0.003 1275
s2 14.05 22 1.17 0 0.105 790
s3 38.4 41.8 1.73 0 0.037 311
s4 39.4 50.2 3 0 0.146 2593
s5 71.1 105.5 1.49 0 0.461 43
s6 16.2 20.3 1.23 1 0.013 845
s7 53.1 59.8 0 1 0.021 1023
s8 72.6 85.6 1.65 1 1.473 1805
s9 13.7 16 1 1 0.116 115
s10 97 100.3 2.66 1 0 56
s11 34.15 37 1.38 2 1.383 233
s12 33 36.7 1.3 2 0.02 74
s13 117.5 120.9 1.83 2 0.682 386
s14 50.5 54.1 0.35 2 0.07 260
s15 38.8 42.8 1.25 2 0.083 84
In Table 9, the attribute Type is used to indicate the category of a stock. The attribute varCD represents the variance of cash dividends of a stock.
STEP 3: Assume the pSize is 10, chromosomes are generated to form the initial population as the following sub-steps:
Sub-step 3.1: Assume K is four. Then, randomly divide thirteen stocks into four the groups to form grouping part. Take C1 as an example, its grouping part is generated as G1:{2, 6, 14}, G2:{1, 3, 10,12}, G3:{4, 8, 9, 11, 13, 15}, G4:{5,7}.
Sub-step 3.2: Calculate the average cash dividend of each group. Take group G1 in C1 as an example, the average cash dividend of G1 is 0.92 (= (1.17+1.23+0.35)/3). In the same way, get the average cash dividend of G1 to G5. The average cash dividends of G1, G2, G3, and G4
are 0.92, 1.54, 1.69 and 0.75.
Sub-step 3.3: Calculate the proportion of average cash dividend of each group. Take G1
as an example, the proportion of average cash dividend of G1 is 0.19 (= 0.92 / (0.92+1.54+1.69+0.75)). In the same way, the proportion of average cash dividend of G1, G2,
G3, and G4 are 0.19, 0.31, 0.34 and 0.15.
Sub-step 3.4 and 3.5: Assume the value of numCom was set at three. Three values were then randomly generated within the range [0, 1]. Assume the generated values are {0.41, 0.2, 0.51}, which are collected as R. Thus, only G2 is selected to put into the candidate portfolio.
Sub-steps 3.6 and 3.7: Generate stock portfolio according to the candidate portfolio. Only G2 is selected to put into the candidate portfolio. Therefore, define that b3 is larger than 0.5 and others are smaller than 0.5. Then, the number of purchased units of each group is randomly generated from the range [0, maxUnit]. Assume maxUnit is 40, the ten initial chromosomes are generated as follows:
C1: {2, 6, 14}, {1, 3, 10, 12}, {4, 8, 9, 11, 13, 15}, {5, 7}, 0.15, 31, 0.74, 39, 0.45, 37, 0.09, 18 C2: {1, 5, 7, 12, 14}, {6, 8, 10}, {3, 4, 15}, {2, 9, 11, 13}, 0.34, 39, 0.75, 28, 0.07, 9, 0.01, 12 C3: {1, 9, 11, 13}, {5, 12}, {3, 4, 6, 14, 15}, {2, 7, 8, 10}, 0.65, 38, 0.46, 9, 0.59, 11, 0.30, 31 C4: {10, 13}, {3, 5, 9, 12, 14}, {2, 4, 6}, {1, 7, 8, 11, 15}, 0.31, 13, 0.30, 6, 0.35, 8, 0.88, 23 C5: {1, 2, 6, 10}, {4, 8, 11, 15}, {3, 7, 9, 12, 13, 14}, {5}, 0.49, 35, 0.80, 18, 0.13, 23, 0.38, 35 C6: {2, 5, 8}, {4, 9, 10, 13}, {1, 6, 11, 14}, {3, 7, 12, 15}, 0.91, 38, 0.33, 10, 0.32, 16, 0.51, 26 C7: {2, 6}, {1, 3, 10, 12}, {4, 7, 8, 9, 11, 13, 14, 15}, {5}, 0.76, 25, 0.28, 31, 0.15, 10, 0.64, 32 C8: {12}, {3, 6, 7, 10, 13}, {2, 4, 9, 11, 14}, {1, 5, 8, 15}, 0.55, 20, 0.65, 19, 0.59, 33, 0.30, 10 C9: {2, 6, 14, 7}, {1, 3, 10, 12}, {4, 8, 9, 11, 13, 15}, {5}, 0.56, 16, 0.91, 10, 0.32, 12, 0.38, 19
C10: {1, 2, 4, 7, 10}, {9, 12}, {3, 5, 8, 11, 13, 15}, {6, 14}, 0.68, 36, 0.09, 16, 0.11, 39, 0.75, 26
STEP 4: Calculate fitness value of each chromosome using the following sub-steps:
Sub-step 4.1: Calculate the risk of investor sentiment (RIS) of each chromosome Cq using the following sub-steps:
Sub-step 4.1.1: Generate the stock portfolios. Take chromosome C1 as an example, according to its group part (G1:{2, 6, 14}, G2:{1, 3, 10,12}, G3:{4, 8, 9, 11, 13, 15}, G4:{5,7}), there are 144 (= 3*4*6*2) possible stock portfolios. All of them are collected in the set SP = {{2, 1, 4, 5}, {2, 1, 4, 7},…, {14, 12, 15, 17}}.
Sub-steps 4.1.2 and 4.1.3: Get the minimal difference of trading price minDTP(si) on stock si, from every transactions which are decided by trading signals. Assume the minDTP(si) of s1, s2, …, s15 are 6.5, 1.8, 10.3, 18.6, 9 , 10.6, 19.8, 9.2, 21.2, 26.5, 10.2, 3.8, 6.7, 8.7, -14.7. Thus, MAXminDTP is -1.8 and MINminDTP is -26.5
Sub-step 4.1.4: Calculate the nomalDTP of each stock. Take s1 as an example, the nomalDTP of s1 is 0.81 (= (-6.5-(-26.5)) / (-1.8-(-26.5))). The nomalDTP value of s1, s2, …, s15
are 0.81, 1, 0.66, 0.32, 0.71, 0.64, 0.27, 0.7, 0.21, 0, 0.66, 0.92, 0.8, 0.72, 0.48.
Sub-steps 4.1.5 and 4.1.6: Take chromosome C1 as an example, its group part is (G1:{2, 6, 14}, G2:{1, 3, 10,12}, G3:{4, 8, 9, 11, 13, 15}, G4:{5,7}), and its stock portfolio part is {0.15, 31, 0.74, 39, 0.45, 37, 0.09, 18}. Take SP1: {2, 1, 4, 5} as an example, the subRIS of SP1 is 31.59 (= 0.81*39). In the same way, calculate all the stock portfolios of C1.
Sub-step 4.1.7: Based on the calculated subRIS of all stock portfolios, the RIS of
chromosome C1 is 36.81. In the same way, calculate the RIS of all chromosomes, and the results are shown in Table 10.
Table 10. The RIS of all chromosomes.
Cq RIS(Cq) Cq RIS (Cq)
C1 31.59 C6 41.23
C2 38.15 C7 24.37
C3 29.62 C8 31.66
C4 30.86 C9 39.83
C5 40.69 C10 35.54
Sub-step 4.2: Calculate portfolio satisfaction of each chromosome Cq using following sub-steps:
Sub-step 4.2.1: The ROI of each stock portfolio is then calculated. Take SP1: {2, 1, 4, 5}
as an example, according to stock portfolio part of chromosome C1: {0.15, 31, 0.74, 39, 0.45, 37, 0.09, 18}, the ROI of SP1 ROI(SP1) is 130.5 (= (20 – 17) * 39+ 0.46 * 39 – 39*(17*0.001425+20*0.001425+20*0.003)).
Sub-step 4.2.2: Calculate the suitability of each stock portfolio. Assume that the predefined maximum number of purchased stocks in the portfolio is 3, the predefined maximum investment capital is 1000, the suitability factor of SP1 suitability(SP1) is 4.51 (ICP(SP1) + PP(SP1)), where ICP(SP1) is 1.51 (= 1000/(17*39)) and PP(SP1) is 3 (= 3/1).
Sub-step 4.2.3 and 4.2.4: The portfolio satisfaction of stock portfolio SP1 subPS(SP1) is 28.94 (=ROI(SP1)/suitability(SP1)). Repeating sub-steps 4.2.1 to 4.2.4, the portfolio satisfactions of all stock portfolios.
Sub-step 4.2.5: Based on the calculated sub portfolio satisfactions of all stock portfolios, the portfolio satisfaction of chromosome C1 is 0.024. In the same way, the portfolio satisfactions of all chromosomes are shown in Table 11.
Table 11. The portfolio satisfactions of all chromosomes.
Cq PS(Cq) Cq PS(Cq)
C1 0.024 C6 0.871
C2 1.621 C7 -0.095
C3 -0.068 C8 0.682
C4 2.373 C9 -0.201
C5 2.856 C10 -0.107
Sub-step 4.3: According to the grouping part of each chromosome, calculate the group balance of each chromosome. Take chromosome C1 as an example, since the group part of C1
is: G1:{2, 6, 14}, G2:{1, 3, 10,12}, G3:{4, 8, 9, 11, 13, 15}, G4:{5,7}, the GB(C1) is 0.569. In the same way, the group balances of all chromosomes are shown in Table 12.
Table 12. The group balances of all chromosomes.
Cq GB(Cq) Cq GB(Cq)
C1 0.569 C6 0.599
C2 0.592 C7 0.494
C3 0.582 C8 0.550
C4 0.575 C9 0.544
C5 0.544 C10 0.552
Sub-step 4.4: Calculate the diversity value of each chromosome. According to the two attributes of stocks, the dissimilarity matrix is calculated and shown in Table 13.
Table 13. The dissimilarity matrix of stocks.
s1 s2 s3 s4 s5 s6 s7 s8 s9 s10 s11 s12 s13 s14 s15
s1 0 1 1 1 1 2 1.67 2 2 2 2 2 2 2 2
s2 1 0 1 1 1 1.33 1.67 2 2 2 2 2 2 2 2
s3 1 1 0 1 0.67 2 2 2 1.33 1.67 1.33 1.67 1.33 1.33 1.67
s4 1 1 1 0 1 2 2 2 2 2 2 2 2 2 2
s5 1 1 0.67 1 0 2 2 2 1.33 1 1.33 1 1.67 1.67 1
s6 2 1.33 2 2 2 0 0.33 1 1 1 2 2 2 2 2
s7 1.67 1.67 2 2 2 0.33 0 1 1 1 2 2 2 2 2
s8 2 2 2 2 2 1 1 0 1 1 2 2 2 2 2
s9 2 2 1.33 2 1.33 1 1 1 0 0.33 1.33 1 1.67 1.33 1
s10 2 2 1.67 2 1 1 1 1 0.33 0 1.33 1 1.67 1.67 1
s11 2 2 1.33 2 1.33 2 2 2 1.33 1.33 0 0.33 0.33 0 0.33
s12 2 2 1.67 2 1 2 2 2 1 1 0.33 0 0.67 0.33 0
s13 2 2 1.33 2 1.67 2 2 2 1.67 1.67 0.33 0.67 0 0.33 0.67
s14 2 2 1.33 2 1.67 2 2 2 1.33 1.67 0 0.33 0.33 0 0.33
s15 2 2 1.67 2 1 2 2 2 1 1 0.33 0 0.67 0.33 0
Take chromosome C1 as an example, its group part is (G1:{2, 6, 14}, G2:{1, 3, 10,12}, G3:{4, 8, 9, 11, 13, 15}, G4:{5,7}), and its stock portfolio part is {0.15, 31, 0.74, 39, 0.45, 37, 0.09, 18}. The diversity value of G1 is 1.777 (= ((dissMatrix(2, 6) + dissMatrix(2, 14) + dissMatrix(6, 14)) / 3 = (1.33+2+2) / 3). In the same way, the diversity value of G2, G3 and G4
are 2.335, 3.722 and 1. Therefore, the DF(C1) is 2.221 (= (1.777+2.335+3.772+1)/4). Similarly, we get the diversity values of all chromosomes and show them in Table 14.
Table 14. The diversity values of all chromosomes.
Cq DF(Cq) Cq DF(Cq)
C1 2.221 C6 2.875
C2 1.813 C7 1.895
C3 2.312 C8 2.139
C4 1.5 C9 2.173
C5 1.486 C10 1.719
Sub-step 4.5: Calculate the fitness value of each chromosome. Assume the parameter and β are set at 1 and 1. Take chromosome C1 as an example, the value of f1(Cq) is 0.958 (=
31.59*0.024*0.569*2.221).
Step 5: Execute selection operation on the population. In this example, the ten initial chromosomes {C1, C2, …,C10} are selected for the next population.
Step 6: Execute crossover operation on the population. Assume chromosomes C1 and C2
are selected randomly. The representation of C1 and C2 follows that C1: {2, 6, 14}, {1, 3, 10, 12}, {4, 8, 9, 11, 13, 15}, {5, 7}, 0.15, 31, 0.74, 39, 0.45, 37, 0.09, 18 and C2: {1, 5, 7, 12, 14}, {6, 8, 10}, {3, 4, 15}, {2, 9, 11, 13}, 0.34, 39, 0.75, 28, 0.07, 9, 0.01, 12. The first phase is for crossover operation on the grouping part. Assume the base chromosome is C1. Let the insertion position of the base chromosome C1 be between G1 and G2, and the insertion sequence of the chromosome C2 be G2: {6, 8, 10}. The group part of C1 becomes G1: {2, 14}, G2: {6, 8, 10}, G3: {1, 3, 4, 9, 12, 13}, G4: { 5, 7, 11, 15}. The second phase crossover operation is on the grouping part. Assume the cut point is one, the new form stock portfolio parts of C1 and C2 are 0.34, 39, 0.74, 39, 0.45, 37, 0.09, 18 and 0.15, 31, 0.75, 28, 0.07, 9, 0.01, 12.
Step 7: Execute mutation operation on the population. Take chromosome C3 as an example.
The representation of C3 follows that C3: {1, 9, 11, 13}, {5, 12}, {3, 4, 6, 14, 15}, {2, 7, 8, 10}, 0.65, 38, 0.46, 9, 0.59, 11, 0.30, 31. In the first phase, it randomly moves a stock from a group to another group. Assume stock s10 in G4 is moved to G2, the grouping part of chromosome C3
becomes G1:{1, 9, 11, 13}, G2:{5, 10, 12}, G3:{3, 4, 6, 14, 15}, G4:{2, 7, 8}. In the second phase, assume b2 of C3 is mutated, the portfolio part of C3 becomes 0.65, 38, 0.82, 9, 0.59, 11, 0.30, 3.
Step 8: Execute inversion operation on the population. Take chromosome C4 as an example. The representation of C4 follows that C4: {10, 13}, {3, 5, 9, 12, 14}, {2, 4, 6}, {1, 7, 8, 11, 15}, 0.31, 13, 0.30, 6, 0.35, 8, 0.88, 23. Assume G1 and G4 are exchanged, the grouping part of chromosome C4 becomes G1: {1, 7, 8, 11, 15}, G2: {3, 5, 9, 12, 14}, G3: {2, 4, 6}, G4: {10, 13}.
Steps 9 and 10: If the stop criterion is satisfied, the group stock portfolio with best fitness value is outputted. After one hundred generations, the group stock portfolio is as follows:
Cbest: G1:{1, 3, 9, 11}, G2:{2, 6, 10}, G3:{4, 5, 15}, G4:{7, 8, 13}, 0.37, 20, 0.84, 9, 0.68, 15, 0.77, 31.
The Cbest represents that it divides fifteen companies {s1, s2, …,s15} into four groups G1, G2, G3 and G4. In grouping part, it shows that group G1 contains s1, s3, s9 and s11; group G2 contains s2, s6 and s10; group G3 contains s4, s5 and s15;and group G4 contains s7, s8 and s13. In stock portfolio part, it shows that stock portfolios can be formed from G2, G3 and G4. Twenty-seven (= 3*3*3) stock portfolios can be generated from Cbest.
CHAPTER 4
Experimental Result
In this section, we use the experiment to show the performance of the proposed approach.
Predefine the parameters in this experiment is as follows: The initial population size Psize was set at 50, the crossover rate pc was set at 0.8, the mutation rate pm was set at 0.03, the inversion rate pi was set at 0.6 and the number of generations G was set at 100. The stock holding day H was set at 3, the maximum number of purchased stocks in the portfolio numCom was set at 3, the predefined maximum investment capital maxInves was set at 1 million, the maximum number of purchased units of a stock maxUnit was set at 40, the parameter was set at 2, and the number of groups K was set at 6. The experimental datasets are described in section 4.1, the analysis of the proposed approach is given in section 4.2, and the comparison of the proposed approach and existing approach is in section 4.3.
4.1 Data Descriptions
In this study, the experimental dataset contains 46 stocks that were collected from the Taiwan Economic Journal (TEJ) database from 2011/01/01 to 2015/12/31. The 46 stocks are selected from FTSE TWSE Taiwan 50 Index. When a company is in FTSE TWSE Taiwan 50 Index, it means that its market capitalization is among the top-50 in the Taiwan securities market. The 46 stock price series are shown in Fig. 4.
Fig. 4. The 46 stock price series.
In Fig. 4, the stock price trend of 46 stocks is shown. Most stocks fluctuate in the range of no more than 150. A small number of stocks fluctuate much even more than 400. The number of stocks’ industrial category is seventeen. In each category of the dataset, the number of stocks, the average of cash dividends and stock symbol are shown in Table 15.
Table 15. The information of all stock categories in dataset.
Stock Category Number of Stocks Avg. Cash
Dividends Stock Symbol
Semiconductor 4 1.124 3474 2330 2311 2325
2408
Table 15 shows that the number of stocks in each category is seventeen. The number of financial stocks is larger than orders in Taiwan 50 Index. The highest Avg. Cash Dividends of categories in Taiwan 50 Index is the car industry.
4.2 Analysis of Group Stock Portfolio
In this section, we analyze the derived diverse GSPs with investor sentiment index by the proposed approach. The training period is from 2011/1/1 to 2013/12/31. The results using formula (24) are shown in Table 16.
Table 16. The derived diverse GSPs with investor sentiment index.
The diverse GSP with investor sentiment index on the dataset.
Group and stock parts Stock portfolio
G1: { 4938, 1326, 2409, 2890, 1102, 2395 } 0.86 5.00 0.61 10.00 0.64 10.00 0.72 39.00 0.18 17.00 0.84 10.00 G2: { 3474, 2311, 2382, 2886 } Fitness Value= 17.06 G3: { 3481, 2303, 2002, 2880, 1216, 2891,
2325, 2105, 2308, 2408, 2301, 2881 } PortfolioSatisfaction= 51.07
G4: { 2207, 2912, 1476 } Group Balance= 2.82
G5: {2892, 2474, 1301, 9904, 2330, 3045,
4904, 2317, 2354, 2884, 6505 } Diversity=5.85 G6: { 2882, 1303, 2883, 1402, 2324, 2801,
1101, 2412, 2887, 2885 } (UB = 1.00, PB = 1.51)
In this example, 46 stocks are divided into 6 groups by this chromosome and 1536 (=
16*12*4*1*2) stock portfolios can be generated for the investors. Then, we focus on the stocks in the same groups. The stock-price series of the derived diverse GSP with investor sentiment index are shown in Fig. 5, Fig. 6, Fig. 7, Fig. 8, Fig. 9 and Fig. 10. In these figures, the horizontal axis represents the trading dates and the vertical axis the stock price.
Fig. 5. The stock-price series in Group1 of the chromosome.
Fig. 6. The stock-price series in Group2 of the chromosome.
Fig. 7. The stock-price series in Group3 of the chromosome.
Fig. 8. The stock-price series in Group4 of the chromosome.
Fig. 9. The stock-price series in Group5 of the chromosome.
Fig. 10. The stock-price series in Group6 of the chromosome.
The Fig. 5 shows that 1326 and 2395 are different from orders in Group1. The Fig. 6 shows that 3474 is different from orders in Group2. The Fig. 7 shows that 2105 and 2308 are different from orders in Group3. The Fig. 8 shows that the stocks is similar in Group4. The Fig.
9 shows that 2474 is different from orders in Group5. The Fig. 10 shows that 1303 and 2412 are different from orders in Group6.
4.3 Comparison of the Proposed Approach and Existing Approach
In this section, experiments show the profits of DGSPs with the investor sentiment index.
For this goal, we compare this proposed approach with the existing approach. Both the proposed approach and the existing approach are used for finding DGSPs. However, the proposed approach contains the concept of investor sentiment and trading strategy, but the existing approach does not. The input data and experimental parameters were set the same in two experiments. Here, we use three standards to compare the profits of approaches, including average ROI, maximum ROI and minimum ROI. The Cbest values executing ten runs by our approach and the previous DGSP are shown in Table 17.
Table 17. The comparison results of two approaches with ten GSPs.
Cbest AvgROI MaxROI MinROI avgAvgROI avgMaxROI avgMinROI C1new 0.336 0.394 0.271 average value of ten MaxROIs, and the avgMinROI represents the average value of MinROIs.
Although the avgMaxROI of the proposed approach is lower than the existing approach, the most important standard avgAvgROI and the avgMinROI of the proposed approach are higher than the existing approach. Finally, we can conclude that diverse GSPs with investor sentiment index can provide a more stable return than the diverse GSPs without investor sentiment index.
CHAPTER 5
Conclusion and Future Work
In this paper, we have proposed an approach for obtaining a diverse group stock portfolio by the group genetic algorithm with investor sentiment index. We add the concept of investor sentiment to the previous approach, and provide the appropriate opportunity for trading stocks to the stock portfolio. In the investor sentiment part, we use the interaction of two investor sentiment indices to figure out the actual investor sentiment. Then, we use the investor sentiment to define the trading strategy. The trading strategy is used to determine the buying timing and selling timing for each stock. With the trading timing, we get the stock prices of all stocks. Based on the previous approach, we design a new factor, the risk of investor sentiment (RIS), related to investor sentiment to enhance the efficiency of finding a diverse group stock portfolio. In the experiment, we show the performance of the proposed approach by comparing the proposed approach with the existing approach. The result shows that the proposed approach can have a better return than the previous approach. It means that the derived diverse group stock portfolio with investor sentiment index can provide a more stable return than the previous approach.
In the future, more investor sentiment indices will join to improve the efficiency of the proposed approach.
References
[1] E. Baralis, L. Cagliero and P. Garza, “Planning stock portfolios by means of weighted frequent itemsets,” Expert Systems with Applications, Vol. 86, No. 15, pp. 1-17, 2017.
[2] G. W. Brown and M. T. Cliff, “Investor sentiment and near term stock market,” Journal of Empirical Finance, Vol. 11, No. 4, pp. 627-628, 2001.
[3] J. Brimberg, N. Mladenović and D. Urošević, “Solving the maximally diverse grouping problem by skewed general variable neighborhood search,” Information Sciences, Vol.
295, pp. 650-675, 2015.
[4] L. Blum and R. L. Rivest, “Training a 3-node neural networks is NP-complete,” Neural Networks, Vol. 5, pp. 117-127, 1992.
[5] J. Bermúdez, J. Segura and E. Vercher, “A multi-objective genetic algorithm for cardinality constrained fuzzy portfolio selection,” Fuzzy Sets and Systems, Vol. 188, pp.
16-26, 2012.
[6] M. Baker and J. Wurgler, “The equity share in new issues and aggregate stock returns,”
Journal of Finance, Vol. 55, No. 5, pp. 2219-2257, 2000.
[7] C. H. Chen and C. Y. Hsieh, “Mining actionable stock portfolio by genetic algorithms,”
Journal of Information Science and Engineering, Vol. 32, No. 6, pp. 1657-1678, 2016.
[8] C. H. Chen, C. B. Lin and C. C. Chen, "Mining group stock portfolio by using grouping genetic algorithms," The IEEE Congress on Evolutionary Computation, pp. 738-743, 2015.
[9] C. H. Chen, C. Y. Lu, T. P. Hong and J. H. Su, “Using grouping genetic algorithm to mine
diverse group stock portfolio,” The IEEE Congress on Evolutionary Computation, pp.
4734-4738, 2016.
[10] J. B. DeLong, A. Shleifer, L. H. Summers and R. J. Waldmann, “Noise trader risk in financial markets,” Journal of Political Economy, Vol. 98, pp. 703-738, 1990.
[11] W. F. M. De-Bondt and R. Thaler, “Dose the stock market overreact?,” The Journal of Finance, Vol. 40, No. 3, pp. 793-805, 1984.
[12] M. Escobar, S. Ferrandoa and A. Rubtsov, “Portfolio choice with stochastic interest rates and learning about stock return predictability,” International Review of Economics &
Finance, Vol. 41, pp. 347-370, 2016.
[13] Z. G. M. Elhachloufi and F. Hamza, “Stocks portfolio optimization using classification and genetic algorithms,” Applied Mathematical Sciences, Vol. 6, pp. 4673-4684, 2012.
[14] E. F. Fama, “Efficient capital markets: a review of theory and empirical work,” The Journal of Finance, Vol. 25, No. 2, pp. 383-417, 1970.
[15] E. Falkenauer, “A New representation and operators for genetic algorithms applied to grouping problems,” Evolutionary Computation, Vol. 2, pp. 123-144, 1994.
[16] E. Falkenauer, “A hybrid grouping genetic algorithm for bin packing,” Journal of Heuristics, Vol. 2, pp. 5-30, 1996.
[17] E. Falkenauer, Genetic Algorithms and Grouping Problems, John Wiley and Sons, 1998.
[18] Z. P. Fan, Y. Chen, J. Ma and S. Zeng, “A hybrid genetic algorithmic approach to the maximally diverse grouping problem,” Journal of the Operational Research Society, Vol.
62, pp. 92-99, 2011.
[19] J. J. Grefenstette, “Optimization of control parameters for genetic algorithms,” IEEE
Transactions on System Man, and Cybernetics, Vol. 16, pp. 122-128, 1986.
[20] D. E. Goldberg, Genetic algorithms in search, optimization, and machine learning, Addison Wesley, 1989.
[21] P. Gupta, M. K. Mehlawat and G. Mittal, “Asset portfolio optimization using support vector machines and real-coded genetic algorithm,” Journal of Global Optimization, Vol.
53, pp. 297-315, 2012.
[22] P. Gupta, M. K. Mehlawat and A. Saxena, “Hybrid optimization models of portfolio selection involving financial and ethical considerations,” Knowledge-Based Systems, Vol.
37, pp. 318-337, 2012.
[23] J. H. Holland, Adaptation in natural and artificial systems, University of Michigan Press, 1975.
[24] G. S. M. Thakur, R. Bhattacharyya and S. Sarkar, “Stock portfolio selection using Dempster–Shafer evidence theory,” Journal of King Saud University - Computer and Information Sciences, pp. 1-13, 2016.
[25] T. P. Hong, C. H. Chen and F. S. Lin, “Using group genetic algorithm to improve performance of attribute clustering,” Applied Soft Computing, Vol. 29, pp. 371-378, 2015.
[26] T. L. James, E. C. Brown and K. B. Keeling, “A hybrid grouping genetic algorithm for the cell formation problem,” Computers & Operations Research, Vol. 34, No. 7, pp. 2059-2079, 2007.
[27] W. Y. Lee, C. X. Jiang and D. C. Indro, “Stock market volatility, excess returns, and the role of investor sentiment,” Journal of Banking & Finance, Vol. 26, No. 12, pp. 2277-2299, 2002.
[28] H. Markowitz, “Portfolio selection,” The Journal of Finance, Vol. 7, No. 1, pp. 77-91,
1952.
[29] P. L. McLeod and S. A. Lobel, “The effects of ethnic diversity on idea generation in small groups,” Academy of Management Proceedings, pp. 227-231, 1992.
[30] H. M. Markowitz, Harry Markowitz: selected works, World Scientific Publishing Company, 2009.
[31] R. Neal and S. M. Wheatley, “Do measures of investor sentiment predict returns,” Journal of Financial and Quantitative Analysis, Vol. 33, No. 4, pp. 523-547, 1998.
[32] G. Pankratz, “A grouping genetic algorithm for the pickup and delivery problem with time windows,” Operations Research Spectrum, Vol. 27, pp. 21-41, 2005.
[33] M. Quiroz-Castellanos, L. Cruz-Reyes, J. Torres-Jimenez, C. Gómez S., H. J. F. Huacuja and A. C. F. Alvim, “A grouping genetic algorithm with controlled gene transmission for the bin packing problem,” Computers & Operations Research, Vol. 55, pp. 52-64, 2015.
[34] B. Rekiek, A. Delchambre and H. A. Saleh, “Handicapped person transportation: An application of the grouping genetic algorithm,” Engineering Application of Artificial Intelligence, Vol. 19, pp. 511-520, 2006.
[35] F. J. Rodriguez, M. Lozano, C. Garcia-Martinez and J. Gonzalez-Barrera, “An artificial bee colony algorithm for the maximally diverse grouping problem,” Information Sciences, Vol. 230, pp. 183-196, 2013.
[36] M. Schmeling, “Institutional and Individual Sentiment: Smart Money and Noise Trader Risk?,” International Journal of Forecasting, Vol. 23, pp. 127-145, 2007.
[37] X. Sun and Z. Liu, “Optimal portfolio strategy with cross-correlation matrix composed by
[37] X. Sun and Z. Liu, “Optimal portfolio strategy with cross-correlation matrix composed by