Introducing neural networks to predict stock prices

Overview

IntroNeuralNetworks in Python: A Template Project

forthebadge made-with-python

GitHub license PRs Welcome

IntroNeuralNetworks is a project that introduces neural networks and illustrates an example of how one can use neural networks to predict stock prices. It is built with the goal of allowing beginners to understand the fundamentals of how neural network models are built and go through the entire workflow of machine learning. This model is in no way sophisticated, so do improve upon this base project in any way.

The core steps involved is: download stock price data from Yahoo Finance, preprocess the dataframes according to specifications for neural network libraries and finally train the neural network model and backtest over historical data.

This model is not meant to be used to live trade stocks with. However, with further extensions, this model can definitely be used to support your trading strategies.

I hope you find this project useful in your journey as a trader or a machine learning engineer. Personally, this is my first major machine learning and python project, so I'll appreciate if you leave a star.

As a disclaimer, this is a purely educational project. Any backtested results do not guarantee performance in live trading. Do live trading at your own risk. This guide and further analysis has been cross-posted in my blog, Engineer Quant

Contents

Overview

The overall workflow for this project is as such:

  1. Acquire the stock price data - this will give us our features for the model.
  2. Preprocess the data - make the train and test datasets.
  3. Use the neural network to learn from the training data.
  4. Backtest the model across a date range.
  5. Make useful stock price predictions
  6. Supplement your trading strategies with the predictions

Although this is very general, it is essentially what you need to build your own machine learning or neural network model.

Getting Started

For those of you that do not want to learn about the construction of the model (although I highly suggest you to), clone and download the project, unzip it to your preferred folder and run the following code in your computer.

pip install -r requirements.txt
python LSTM_model.py

It's as simple as that!

Requirements

For those who want a more details manual, this program is built in Python 3.6. If you are using an earlier version of Python, like Python 3.x, you will run into problems with syntax when it comes to f strings. I do suggest that you update to Python 3.6.

pip install -r requirements.txt

Stock Price Data

Now we come to the most dreaded part of any machine learning project: data acquisiton and data preprocessing. As tedious and hard as it might be, it is vital to have high quality data to feed into your model. As the saying goes "Garbage in. Garbage out." This is most applicable to machine learning models, as your model is only as good as the data it is fed. Processing the data comes in two parts: downloading the data, and forming our datasets for the model. Thanks to Yahoo Finance API, downloading the stock price data is relatively simple (sadly I doubt not for long).

To download the stock price data, we use pandas_datareader which after a while did not work. So we use this fix and use fix_yahoo_finance. If this fails (maybe in the near future), you can just download the stock data directly from Yahoo for free and save it as stock_price.csv.

Preprocessing

Once we have the stock price data for the stocks we are going to predict, we now need to create the training and testing datasets.

Preparing Train Dataset

The goal for our training dataset is to have rows of a given length (the number of prices used to predict) along with the correct prediction to evaluate our model against. I have given the user the option of choosing how much of the stock price data you want to use for your training data when calling the Preprocessing class. Generating the training data is done quite simply using numpy.arrays and a for loop. You can perform this by running:

Preprocessing.get_train(seq_len)

Preparing Test Dataset

The test dataset is prepared in precisely the same way as the training dataset, just that the length of the data is different. This is done with the following code:

Preprocessing.get_test(seq_len)

Neural Network Models

Since the main goal of this project is to get acquainted with machine learning and neural networks, I will explain what models I have used and why they may be efficient in predicting stock prices. If you want a more detailed explanation of neural networks, check out my blog.

Multilayer Perceptron Model

A multilayer perceptron is the most basic of neural networks that uses backpropagation to learn from the training dataset. If you want more details about how the multilayer perceptron works, do read this article.

LSTM Model

The benefit of using a Long Short Term Memory neural network is that there is an extra element of long term memory, where the neural network has data about the data in prior layers as a 'memory' which allows the model to find the relationships between the data itself and between the data and output. Again for more details, please read this article

Backtesting

My backtest system is simple in the sense that it only evaluates how well the model predicts the stock price. It does not actually consider how to trade based on these predictions (that is the topic of developing trading strategies using this model). To run just the backtesting, you will need to run

back_test(strategy, seq_len, ticker, start_date, end_date, dim)

The dim variable is the dimensions of the data set you want and it is necessary to successfully train the models.

Stock Predictions

Now that your model has been trained and backtested, we can use it to make stock price predictions. In order to make stock price predictions, you need to download the current data and use the predict method of keras module. Run the following code after training and backtesting the model:

data = pdr.get_data_yahoo("AAPL", "2017-12-19", "2018-01-03")
stock = data["Adj Close"]
X_predict = np.array(stock).reshape((1, 10)) / 200
print(model.predict(X_predict)*200)

Extensions

As mentioned before, this projected is highly extendable, and here some ideas for improving the project.

Getting Data

Getting data is pretty standard using Yahoo Finance. However, you may want to look into clustering data in terms of trends of stocks (maybe by sector, or if you want to be really precise, use k-means clustering?).

Neural Network Model

This neural network can be improved in many ways:

  1. Tuning hyperparameters: find the optimal hyperparameters that gives the best prediction
  2. Backtesting: Make the backtesting system more robust (I have left certain important aspects out for you to figure). Maybe include buying and shorting?
  3. Try different Neural Networks: There are plenty of options and see which works best for your stocks.

Supporting Trade

As I said earlier, this model can be used to support trading by using this prediction in your trading strategy. Examples include:

  1. Simple long short strategy: you buy if the prediction is higher, and vice versa
  2. Intraday Trading: if you can get your hands on minute data or even tick data, you can use this predictor to trade.
  3. Statistical Arbitrage: use can also use the predictions of various stock prices to find the correlation between stocks.

Contributing

Feel free to fork this and submit PRs. I am open and grateful for any suggestions or bug fixes. Hope you enjoy this project!


For more content like this, check out my academic blog at https://medium.com/engineer-quant

Owner
Vivek Palaniappan
Keen on finding effective solutions to complex problems - looking into the broad intersection between engineering, finance and AI.
Vivek Palaniappan
QR2Pass-project - A proof of concept for an alternative (passwordless) authentication system to a web server

QR2Pass This is a proof of concept for an alternative (passwordless) authenticat

4 Dec 09, 2022
This is the official source code of "BiCAT: Bi-Chronological Augmentation of Transformer for Sequential Recommendation".

BiCAT This is our TensorFlow implementation for the paper: "BiCAT: Sequential Recommendation with Bidirectional Chronological Augmentation of Transfor

John 15 Dec 06, 2022
Self-Supervised Image Denoising via Iterative Data Refinement

Self-Supervised Image Denoising via Iterative Data Refinement Yi Zhang1, Dasong Li1, Ka Lung Law2, Xiaogang Wang1, Hongwei Qin2, Hongsheng Li1 1CUHK-S

Zhang Yi 72 Jan 01, 2023
Implementation based on Paper - Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

Implementation based on Paper - Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

HamasKhan 3 Jul 08, 2022
[CVPR 2022] Structured Sparse R-CNN for Direct Scene Graph Generation

Structured Sparse R-CNN for Direct Scene Graph Generation Our paper Structured Sparse R-CNN for Direct Scene Graph Generation has been accepted by CVP

Multimedia Computing Group, Nanjing University 44 Dec 23, 2022
This is the official implementation of the paper "Object Propagation via Inter-Frame Attentions for Temporally Stable Video Instance Segmentation".

ObjProp Introduction This is the official implementation of the paper "Object Propagation via Inter-Frame Attentions for Temporally Stable Video Insta

Anirudh S Chakravarthy 6 May 03, 2022
Image De-raining Using a Conditional Generative Adversarial Network

Image De-raining Using a Conditional Generative Adversarial Network [Paper Link] [Project Page] He Zhang, Vishwanath Sindagi, Vishal M. Patel In this

He Zhang 216 Dec 18, 2022
This is a repo of basic Machine Learning!

Basic Machine Learning This repository contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles and other resource

Ekram Asif 53 Dec 31, 2022
Scikit-learn compatible estimation of general graphical models

skggm : Gaussian graphical models using the scikit-learn API In the last decade, learning networks that encode conditional independence relationships

213 Jan 02, 2023
HyperLib: Deep learning in the Hyperbolic space

HyperLib: Deep learning in the Hyperbolic space Background This library implements common Neural Network components in the hypberbolic space (using th

105 Dec 25, 2022
This repository builds a basic vision transformer from scratch so that one beginner can understand the theory of vision transformer.

vision-transformer-from-scratch This repository includes several kinds of vision transformers from scratch so that one beginner can understand the the

1 Dec 24, 2021
a general-purpose Transformer based vision backbone

Swin Transformer By Ze Liu*, Yutong Lin*, Yue Cao*, Han Hu*, Yixuan Wei, Zheng Zhang, Stephen Lin and Baining Guo. This repo is the official implement

Microsoft 9.9k Jan 08, 2023
[ICCV'21] Pri3D: Can 3D Priors Help 2D Representation Learning?

Pri3D: Can 3D Priors Help 2D Representation Learning? [ICCV 2021] Pri3D leverages 3D priors for downstream 2D image understanding tasks: during pre-tr

Ji Hou 124 Jan 06, 2023
Tools for computational pathology

A toolkit for computational pathology and machine learning. View documentation Please cite our paper Installation There are several ways to install Pa

254 Dec 12, 2022
Facebook Research 605 Jan 02, 2023
Point detection through multi-instance deep heatmap regression for sutures in endoscopy

Suture detection PyTorch This repo contains the reference implementation of suture detection model in PyTorch for the paper Point detection through mu

artificial intelligence in the area of cardiovascular healthcare 3 Jul 16, 2022
Azion the best solution of Edge Computing in the world.

Azion Edge Function docker action Create or update an Edge Functions on Azion Edge Nodes. The domain name is the key for decision to a create or updat

8 Jul 16, 2022
GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration

GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration Stefan Abi-Karam*, Yuqi He*, Rishov Sarkar*, Lakshmi Sathidevi, Zihang Qiao, Co

Sharc-Lab 19 Dec 15, 2022
an implementation of Video Frame Interpolation via Adaptive Separable Convolution using PyTorch

This work has now been superseded by: https://github.com/sniklaus/revisiting-sepconv sepconv-slomo This is a reference implementation of Video Frame I

Simon Niklaus 985 Jan 08, 2023
Implementation of Pooling by Sliced-Wasserstein Embedding (NeurIPS 2021)

PSWE: Pooling by Sliced-Wasserstein Embedding (NeurIPS 2021) PSWE is a permutation-invariant feature aggregation/pooling method based on sliced-Wasser

Navid Naderializadeh 3 May 06, 2022