Beta preview of a forecast service for the Swedish OMX stock market.

To create trading strategies the power of deep learning models, Elman neural networks and other algorithms are used which is a very computationally intensive task.To make it work I built an architecture of clustered servers that dispatch incoming requests from the web front-end to a back-end of statistical computing resources. Now in the beta preview, the service is a bit underpowered in terms of CPU and memory, so users can expect lag and It may take a little while to load charts.

First, the shares are ranked on OMX Stockholm Small, Mid and Large cap based on averaging the 2, 4, and 6-month returns and are presented in a searchable drop list with the shares with the best returns at the top of the list. When a stock is selected from the drop down list, a back test is performed between the dates selected. The back-test uses a strategy that consists of a combination of several technical indicators: Bollinger bands, CCI, EMA and SAR. Two different rules are used for buying, one to find the reversal after a downtrend and the second to find buy signals in an up-trend. A rule is used to find the time when to sell the stock. A stop trailing order are placed directly after a share has been bought. The rule will sell the stock if it passes the stop trailing order percent. The Backtest tab shows the stock chart with the opening price and a list of actions performed by the backtest: Sell and the profit or loss. The actions buy or sell are presented one day in advance and are executed on the next day opening price.