Project for time series forecasting with machine learning.

Cleaned dataset changes:

• zeros filled with previous prices for price dfs
• duplicate entries removed(all previous prices = current prices)

Features:

• Previous buy prices up to stepsBack
• Current buy quantity
• Current capital flow (cf = currentPrice * currentQuantity)

Layers:

• encoder(2 LSTM)
• decoder(2 LSTM)(teacher forcing)
• timeDistributed(Dense)

Loss:

• Huber
• mse
• custom for multiple outputs

Credits to chriskok for the original data and basic framework for this code. https://github.com/chriskok/GEPrediction-OSRS