Machine Learning and Deep Learning for time series forecasting

This repository contains codes, resources and models for time series forecasting and analysis using Machine Learning and Deep Learning

Models

• Density Hemisphere Neural Network (DensityHNN): Implements the model proposed in the Paper : From Reactive to Proactive Volatility Modeling with Hemisphere Neural Networks.

DensityHNN is a deep Learning algorithm designed to produce Density forecasts on time-dependent data by using an ensemble of deep neural networks. The proposed architecture is the following:

.

The network has two independent hemispheres: one estimating the conditional mean (yellow) and one estimating the conditional volatility (blue). Both hemispheres share a common block (red) at the entrance of the network, which performs a non-linear transformation of covariates before sending them to the two hemispheres.

After estimation, the model is capable of producing conditional forecasts along with uncertainty estimates.

.

A simple usage tutorial for the density hemisphere neural network is available here example.

Getting Started

1. Clone this repository:

   git clone https://github.com/TheAionxGit/aionx.git

2. Install with pip

   pip install aionx

3. (TODO) explore the example notebooks in the Link to Tutorial Notebook directory to get started.

Dependencies

• NumPy : The fundamental package for scientific computing in Python.
• Pandas : An open-source data analysis and manipulation library.
• TensorFlow : An open-source deep learning framework.
• Scikit-Learn : An open-source machine learning framework.