Neural networks can discover patterns in data that humans might not notice and successfully predict the future trend. Addaptron Software has developed NNSTP-2, neural network computer tool, to help stock traders in predicting stock prices for short terms. NNSTP-2 predicts future share prices or their percentage changes (can be chosen in settings menu) using Fuzzy Neural Network (FNN). It operates automatically when creating the FNN, training it, and mapping to classify a new input vector.
Recommended forecasting horizon is within the range of one to around 60 trading days. The software predicts closing price or weighted one (can be chosen in settings menu). Input data are weighted closing price of the stock and the volume traded (EOD csv-files). The historical period should be multi-fold more than forecast period. The maximum number of predicted days is defined by the formula: (predicted days) = [(historical days) - 40] / 20. The input data transformed to characteristic matrices before training FNN. The forecasting is based on automatic scan of different inputs periods to define accuracy of each one by back testing. Then the final forecast is built on weighted averaging of all forecasts. Each weight is proportional to the accuracy of a certain input period forecast.
NNSTP-2 has a user-friendly easy-to-use interface. The software is intended for traders with a basic knowledge in stock market analysis.
NNSTP-2 is a part of integrated system SMFT-1 (starting from 2011)
Some additional info is available on Q & A page. The detailed description is presented in User's Manual (accessible from menu Help after downloading and installing SMFT-1).
The algorithm of using NNSTP-2 is simple - select input data, run forecast, and view result. The input data are historical quotes of the shares selected by user. The output result (forecast) is presented in a chart form and data table (in text-box and text file).