Using custom built ML Algorithms from scratch and combining it with the power of deep learning . This model tries to predict the stock market trend with a mean accuracy of 85% and above for a testing of over a year's data
down.py - Stock market File Downloader
tester.py - tester file for all the models and algorithm which returns a detailed report - SAVED in ./report
run.py - module to return predictions and save it in ./report for all model
./testfilesNS - downloaded NSE stock files.
./del - testing and training sets.
./report - detailed reports from model.
all the algorithms used in the project are custom built for stock market data from scratch . Algorithms USED:
KNN - 45% - 60% accuracy,highly unreliable .
SVM - 67% accuracy at best .
NAIVE BAYES - 55% to 60% accuracy .
Ingestion Engine takes raw data from downloaded stock files and apply proprietary data manipulations to enchance its accuracy by - 11-20%
this data is fed as input for the model for prediction and testing .
MODEL 1 - mixup() - this model combines individual ML algorithms with neural net . But it has a redundancy error from all the ML algorithms,
accuracy of this model is 70% which is basically a pessimistic model i.e. - NO FLEXIBLITY in decisions .
MODEL 2 - this model is an upgrade to model 1 , here the redundancy error is removed , gives a mean accuracy of 89.7 %. the model is quite flexible to predict loss and profit of the stock market .
Prediction Analysis for a stock:
PS : the implementation files for the model and algorithms are missing in this repository on purpose.For further info leave a message at chandangm33@gmail.com