For many people, there is often petty cash on hand that they can do some investment, hoping to gain another way of income. However, the general retail investors usually do not have enough time and resources to conduct a detailed study on the stocks. As a result, most of the retail investors are always unable to make profits when investing in stocks.

　In Taiwan, there are more than 100 funds covering index funds and mutual funds. Each mutual fund relies on a professional investment manager to make good trading decisions for clients. It is possible to purchase an entire mutual fund as an investment, and frequent investment in mutual fund is costly because transaction costs are high. Therefore, direct investment in stocks is a better practice to most retail investors. For Taiwan stocks, many of them can buy one thousand shares or below after the market close, which also makes investment in stocks more flexible and convenient.

　Through researching funds, it can be found that the total monthly return rates of certain funds are always stable. Therefore, if we can predict these profitable funds’ holdings next month, and then invest in similar targets, we will have more stable investment results for retail investors. This study uses the funds’ holdings information provided by the Taiwan Economic Journal(TEJ) to study the changes in the funds’ holdings preferences from 2016 to 2017 for a total of 24 months to predict and verify the ideal portfolio for each month of 2018. In conjunction with the content of the time series, a Long Short-Term Memory neural network (LSTM) is used to build the model. In addition, the use of Matrix Factorization can make the selected targets more diverse while achieving the effect of dispersing risks. According to the experimental results of this study, the forecasted portfolio had a total of 34% higher than the annualized return rate of the Taiwan Index(TAIEX) for the year of 2018.

　For many people, there is often petty cash on hand that they can do some investment, hoping to gain another way of income. However, the general retail investors usually do not have enough time and resources to conduct a detailed study on the stocks. As a result, most of the retail investors are always unable to make profits when investing in stocks.

　In Taiwan, there are more than 100 funds covering index funds and mutual funds. Each mutual fund relies on a professional investment manager to make good trading decisions for clients. It is possible to purchase an entire mutual fund as an investment, and frequent investment in mutual fund is costly because transaction costs are high. Therefore, direct investment in stocks is a better practice to most retail investors. For Taiwan stocks, many of them can buy one thousand shares or below after the market close, which also makes investment in stocks more flexible and convenient.

　Through researching funds, it can be found that the total monthly return rates of certain funds are always stable. Therefore, if we can predict these profitable funds’ holdings next month, and then invest in similar targets, we will have more stable investment results for retail investors. This study uses the funds’ holdings information provided by the Taiwan Economic Journal(TEJ) to study the changes in the funds’ holdings preferences from 2016 to 2017 for a total of 24 months to predict and verify the ideal portfolio for each month of 2018. In conjunction with the content of the time series, a Long Short-Term Memory neural network (LSTM) is used to build the model. In addition, the use of Matrix Factorization can make the selected targets more diverse while achieving the effect of dispersing risks. According to the experimental results of this study, the forecasted portfolio had a total of 34% higher than the annualized return rate of the Taiwan Index(TAIEX) for the year of 2018.

[1] R. Gencay. (1996). Non-linear Prediction of Security Returns with Moving Average Rules. Journal of Forecasting, 15(3), 165-174.

[2] N. Baba & M. Kozaki. (1992). An intelligent forecasting system of stock price using neural networks. IJCNN-92, Vol. 1, 371-377.

[3] Robert P. Schumaker & Hsinchun Chen. (2006). Textual Analysis of Stock Market Prediction Using Financial News Articles. 12th Americas Conference on Information Systems (AMCIS), 1-29.

[4] Sushil Bikhchandani & Sunil Sharma. (2000). Herd behavior in financial markets. IMF Staff Papers, 47(3), 279-280.

[5] Eric G. Falkenstein. (1996). Preferences for Stock Characteristics As Revealed by Mutual Fund Portfolio Holdings. The Journal of Finance, 51(1), 111-135.

[6] James Davidson, Palash Nandy, Benjamin Liebald, Taylor Van Vleet, Junning Liu. (2010). The YouTube video recommendation system. Proceedings of the 2010 ACM Conference on Recommender Systems, RecSys 2010, 293-296.

[7] Yehuda Koren, Robert Bell, Chris Volinsky. (2009). Matrix Factorization Techniques for Recommender Systems. IEEE, Computer Journal, 42(8), 30-37.

[8] Greg Linden, Brent Smith, Jeremy York. (2003). Amazon.com Recommendations: Item-to-Item Collaborative Filtering. IEEE, Internet Computing, 7(1), 76-80.

[9] Abhinandan Das, Mayur Datar, Ashutosh Garg, Shyamsundar Rajaram. (2007). Google news personalization: scalable online collaborative filtering. Proceedings of the 16th International Conference on World Wide Web, WWW 2007, 271-280.

[10 ] Jun Wang, Arjen P. de Vries, Marcel J. T. Reinders. (2006). Unifying user-based and item-based collaborative filtering approaches by similarity fusion. SIGIR conference on Research and development in information retrieval, 501-508.

[11] Badrul Sarwar, George Karypis, Joseph Konstan, John Riedl. (2001). Item-based collaborative filtering recommendation algorithms. Proceedings of the 10th international conference on World Wide Web, 285-295.

[12] Michael J. Pazzani & Daniel Billsus. (2007). Content-Based Recommendation Systems. The Adaptive Web, 325-341.

[13] Xiaoyuan Su & Taghi M. Khoshgoftaar. (2009). A Survey of Collaborative Filtering Techniques. Advances in Artificial Intelligence, Volume 2009, Article ID 421425, 19 pages.

[14] Andrew I. Schein, Alexandrin Popescul, Lyle H. Ungar, David M. Pennock. (2002). Methods and metrics for cold-start recommendations. Proceedings of the 25th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2002), 253-260.

[15] Yoshio Takane, Forrest W. Young, Jan de Leeuw. (1977). Nonmetric individual differences multidimensional scaling: An alternating least squares method with optimal scaling features. Psychometrika, 42(1), 7–67.

[16] Warren S. Sarle. (1995). Stopped training and other remedies for overfitting. Proceedings of the 27th Symposium on the Interface of Computing Science and Statistics, 352-360.

[17] D. E. Rumelhart, G. E. Hinton, R. J. Williams. (1986). Learning internal representations by error propagation. Parallel distributed processing: explorations in the microstructure of cognition, vol. 1, 318-362.

[18] L.K. Hansen & P. Salamon. (1990). Neural Network Ensembles. IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(10), 993-1001.

[19] Sepp Hochreiter & Jurgen Schmidhuber. (1997). LONG SHORT-TERM MEMORY. Neural Computation, 9(8), 1735-1780.

[20] Hasim Sak, Andrew Senior, Francoise Beaufays. (2014). Long Short-Term Memory Based Recurrent Neural Network Architectures for Large Vocabulary Speech Recognition. ArXiv 2014, 1-3.

[21] John Platt. (1991). A resource-allocating network for function interpolation. Neural Computation, 3(2), 213-225.

[22] David S. Scharfstein & Jeremy C. Stein. (1990). Herd Behavior and Investment. The American Economic Review, 80(3), 465-479.

[23] Xiangnan He, Lizi Liao, Hanwang Zhang, Liqiang Nie, Xia Hu, Tat-Seng Chua. (2017). Neural Collaborative Filtering. Proceedings of the 26th International Conference on World Wide Web, 173-182.

[24] Oleksii Kuchaiev & Boris Ginsburg. (2017). Factorization tricks for LSTM networks. ICLR 2017, arXiv:1703.10722v3, 1-4.