Archive for November, 2010

We were wondering whether the batch training (train()) or the online training (adapt()) is the better method to train the ANN. Here is an excerpt from a Matlab forum: ” I would like to know what the differences are between the ADAPT and TRAIN functions. Solution: The difference between these training methods is that ADAPT […]


Before moving fully to the 2 inputs ANN case (dayOfTheWeek, currDayChange), we regress back to a strategy we have already studied before without real success. Our input this time has only the currDayChange and we try to predict the nextDayChange. This is the classical daily Mean Reversion (MR) or Follow Through (FT) strategy. (Prediction of […]


As it was described in the previous post, the performance, representing the dayOfTheWeek input as 5 dimensional data was not better than the best 1 dimensional representation, but was better than the average 1 dimensional representation. Therefore we forsake some performance for stability and we preferred the 5 dimensional input. We mention here this, because […]


We are already well aware the importance of normalizing the output. (See previous posts.) However, we trusted the Matlab FFNN (FeedForward Neural Network) to use the default ‘mapminmax’ input preprocessing to map the input to -1..+1, and therefore we thought it doesn’t really matter how we select our input range, because it will be mapped […]


Let’s refresh our thoughts on the day-of-the-week anomaly. Here is the collection of studies on the weekend effect: http://calendar-effects.behaviouralfinance.net/weekend-effect/ ” The weekend effect (also known as the Monday effect, the day-of-the-week effect or the Monday seasonal) refers to the tendency of stocks to exhibit relatively large returns on Fridays compared to those on Mondays. This […]


The first stage of the NeuralSniffer development is over. We gained significant experience how to train and use the different ANN algorithms. At this stage, we kept everything as simple as we could. (To facilitate our learning and understanding the ANN concept.) We tried not to chase performance. CAGR% or TR% were not important. Using […]


One point of using an ensemble is to decrease randomness. Especially in the case of a homogeneous ensemble. The average of the backtests remains the same as in the standalone case, but the variance decreases. However, another motivation for the ensemble forecast is to increase the performance, increase the average. This is reported by many […]