Forecasting Nifty move

I recently took up a few courses on machine learning and python, so thought of doing some practice on my new skills and hence decided to work on Nifty50 dataset.

The task I took up was to forecast the move of Nifty50 index using deep learning. I know it is childish of me to predict the move of such a dynamic index and I might not get any meaningful results however, it is always fun to try out things than just have the theoretical knowledge.

Approach 1 - Using simple ANN with multiple features

Steps:

  1. Merge Nifty50 datasets - Index Price, Volume, PE, PB, VIX, FII & DII activity
  2. Clean up the master dataset
  3. Identify key features from the master dataset
  4. Split the data into training and test sets
  5. Apply feature scaling
  6. Build artificial neural network using TensorFlow and train the model
  7. Evaluate the model and predict the outcome
  8. Refine the model and reiterate  <-- This step is in the loop and the loop is not over yet :)

Libraries used:
  • numpy, pandas, tensorflow, sklearn, matplotlib, seaborn

So far the results are not much encouraging. However, I am planning to continue working on it and refine the model. 



Approach 2 - Running LTSM on time series index value 

Model Architecture

Model: "sequential"
_________________________________________________________________
 Layer (type)                       Output Shape              Param #   
=================================================================
 lstm (LSTM)                      (None, 60, 64)            16896     
 dropout (Dropout)           (None, 60, 64)            0         
 lstm_1 (LSTM)                  (None, 60, 64)            33024     
 dropout_1 (Dropout)       (None, 60, 64)            0         
 lstm_2 (LSTM)                 (None, 60, 64)            33024     
 dropout_2 (Dropout)      (None, 60, 64)            0         
 lstm_3 (LSTM)                 (None, 64)                   33024     
 dropout_3 (Dropout)      (None, 64)                   0         
 dense (Dense)                   (None, 1)                      65        
=================================================================

Output:



Observations:
Looks like the stip drop due to Covid and sudden v shape recovery followed by it is affecting model prediction.


P.S.: Comments, suggestions, and collaborations are welcome :) 

Comments