Stock Trend Prediction using SV-kNNC and SOM

In the capital market, a stockbroker often finds it difficult to decide whether we should buy or sell the stock. The lengthy time for stock price analysis and the enormous amount of stock data are the two main reasons which cause difficulty. The purpose of this research is the proposed hybrid prediction model for stock trend prediction. The prediction will let the user know whether the stock price will go up or down. It can be used to assist the stockbroker by reducing the analysis time and simplifying the analysis process. The dataset is consists of five blue-chip stocks in the Indonesian Stock Exchange. The model proposed is a modified SV-kNNC (Support Vector - k Nearest Neighbor Clustering) by replacing the K-Means clustering technique with SOM (Self Organizing Map). Our proposed model is short-term stock trend prediction and its performance will be measured by f-measure. The experiments demonstrate that SV-kNNC + SOM is fairly effective model to predict stock trend to go up or down.