Stock Market Prediction & Clustering

The stock market is one of the most dynamic and volatile sources of data.The data is generated every second as the money never sleeps. The change in the values of stocks is a result of the change in the market and sentiments of the people.

In this project, we’ve tried to predict the stocks of National Stock Exchange (NSE) India in Top gainers and Losers over a period of time. Which will be used in clustering the stocks.

This is a python based project and uses Machine learning to predict the value of the stock for the next day.

The Data

Since the data we’ve considered is from NSE, we’re using nsepy and nsetools library to obtain the data. nsepy provides the day-to-day data for the provided symbol.

NSE official website

NSE SYMBOLS list

Taking 4 companies for prediction

1. Brigade Group
2. Godrej Properties Limited
3. ITC Limited
4. Reliance Industries Limited

Preprocessing

The model uses mainly 4 features for prediction — OPEN, HIGH, LOW, CLOSE. Therefore, we’re dropping the extra features present in the DataFrames.

2 features are added to the DataFrames, which provides insight.

1. Change- which represents the change or volatility in the stock, every day.
2. Gain- Categorical feature representing gain or loss in day-to-day value.

Once the extra features are dropped, we have the df.head() and df.describe() of the listed stocks, providing the first 5 values and statistical description of DataFrames, respectively.

Brigade Enterprises LimitedGodrej Properties LimitedITC LimitedReliance Industries LimitedVisualization

Visualizing the features over the years, which gives us information about the company’s past trends and volatility.

Also, included OHLC, HLC and Closing value graph over the years, which is used to illustrate movements in the price of a financial instrument over time

Brigade Group

OHLC and HLC Values over the years

Godrej Properties Limited

OHLC and HLC Values over the years

ITC Limited

OHLC and HLC Values over the years

Reliance Industries Limited

OHLC and HLC Values over the yearsPrediction

The prediction is done using lstm (Long Short Term Memory networks ) because they are pretty good at extracting features in time series data.

The Train-Test split is done in 75–25 % ratio.

For a total 5 cycle for each stock.

After the training for each stock individually, the model predicts on the test dataset and computes the RMSE (Root Mean Square Error).

We get the following results for the same.

Brigade Group

RMSE on test dataset 5.13

Godrej Properties Limited

RMSE on test dataset 14.72

ITC Limited

RMSE on test dataset 6.38

Reliance Industries Limited

RMSE on test dataset 29.19

This post includes one of the modules of the complete project, which is to predict using neural networks. Another aspect of the project includes the sentiment part of the project about the company stock.

When combined the result of both module we can cluster the stocks in gain or lose.

For the complete code, please visit my github repository:

sahilverma0696/Stock-market-prediction-clustering

Stock Market Prediction & Clustering was originally published in Hacker Noon on Medium, where people are continuing the conversation by highlighting and responding to this story.