Introducing ProsperNN-a Python package for forecasting with neural networks

Area of Science:

• Computer Science
• Machine Learning
• Data Science

Background:

• Time series forecasting is crucial for demand and macroeconomic predictions.
• Existing recurrent neural network (RNN) analysis tools lack accessibility and comprehensive features.
• Previous uncertainty visualization methods were limited to specific platforms.

Purpose of the Study:

• Introduce the prosper_nn Python package for advanced time series forecasting.
• Provide accessible implementations of proven neural network models and novel analytical tools.
• Facilitate benchmarking and further development of forecasting techniques.

Main Methods:

• Implementation of four distinct neural network architectures in PyTorch.
• Development of the first sensitivity analysis method tailored for RNNs.
• Integration of a heatmap for visualizing forecasting uncertainty.

Main Results:

• Public release of the prosper_nn package on GitHub.
• Availability of industry-tested models and methods for broader research use.
• Enabling new research avenues in time series analysis and forecasting.

Conclusions:

• The prosper_nn package democratizes access to advanced forecasting tools.
• It empowers researchers and practitioners to enhance time series prediction accuracy and reliability.
• Facilitates wider adoption of sophisticated forecasting methods in various domains.