Image generation, rounding TA Results

omerbsezer / CNN-TA

Algorithmic Financial Trading with Deep Convolutional Neural Networks: Time Series to Image Conversion Approach: A novel algorithmic trading model CNN-TA using a 2-D convolutional neural network based on image processing properties.

115 stars 40 forks source link

Image generation, rounding TA Results #3

Open jerryfane opened 3 years ago

jerryfane commented 3 years ago

In step 1, you transformed the results of the technical indicators by dividing or multiplying them by some constants and rounding them to 2 decimal places.

In the paper you don't explain the motivation for this, is it an important step to get your results back? Why did you do this rather than leaving the values as they were before?

eg.

resultRSI= Precision.round(((rsiArray[5].getValue(counterRow-1).toDouble()**/50**)-1),2) + ";" + etc.

omerbsezer commented 3 years ago

Each technical indicator is normalized in itself. For example; while the RSI varies between [0-100], it is normalized to [-1,1]. The main purpose of this approach is normalizing each indicator to [-1,1]. Rounding to 2-digit decimal is not so important, it could be possible that rounding to 4-digit decimal.