Trust Based Stock Recommendation System – A Social Network Analysis Approach

The CMAS Lab team has  published a research papers on "  Application of Social network analysis in stock market analysis " . 

In this paper , we propose a novel system to recommend the leading investment options in stocks utilizing a methodology based on the transactions of trusted mutual funds and their corresponding stock holding portfolio. The network formed by the stock holding portfolio is the mutual funds are analyised by the tools of social network analysis.

      " The hidden relationships among stocks have greater influence on the degree of accuracy of reliability of portfolio recommendation which in particular have a significant impact on the short-term profitability."

We have carried out an detailed experimentation basing on CRISIL (most trusted external rating agency in India) Rank-1 Indian mutual funds. For this purpose we have collected data from 17 mutual funds which have CRISIL rating of 1, as on March 2014 (quarter ending). The stock investment portfolio covers a wide range 288 stocks from varied sectors having greater than 792 investments (edges in the corresponding graph). 

The primary goal of this experiment is to assess the effectiveness of the proposed recommendation 

system for duration of a financial quarter (01 April 2014 to 30 June 2014). After the portfolio information is extracted, the 2-mode and 1-mode networks were formally constructed using the software packages Pajek  and Gephi11,12. The 2-mode network of mutual funds and stocks is given in Fig. 3(a) and the derived 1-mode network of stocks is plotted in Fig. 3(b). 

 The top 10 stocks based on degree centrality (Portfolio-1) is given in Table  along with corresponding Nifty index in last row for comparison. 

1) closing price on 1 April 2014 which is considered as the buying price.

2) closing price on 30 June 2014, which is considered as the selling price 

3) low and high value during this period. 

Experiment demonstrates that Portfolio-1 gives average normal gain of 20.51 % to its investors. In order to assess the reliability of this  portfolio recommendations we compared them with the performance of Nifty index during the same period and found that performances of this o portfolio is better than the overall performance of Nifty index during the same period reiterating the efficiency of the proposed recommendation framework. 

 In short, it can be concluded safely that this methodology is a reliable portfolio recommendation system for a novice investor who is seeking well-informed investment guidance and this captures investment similarities exist between mutual funds.

The research paper "Trust Based Stock Recommendation System – A Social Network Analysis Approach" is presented in ICICT 2014 and  proceedings of ICICT 2014 available in  Elsevier Procedia Computer Science Volume 46 (2015).
http://www.sciencedirect.com/science/article/pii/S1877050915000885