LSTM vs. ARIMA in Predicting Neurological Trends

Overview

This initiative aims to evaluate the efficacy of LSTM (Long Short-Term Memory) networks versus ARIMA (AutoRegressive Integrated Moving Average) models in forecasting neurological activity, using data collected via a Muse headset. Despite resource constraints, this project makes strides in demonstrating the potential of LSTM models for complex, time-series analysis in neurological data.

Introduction

The prediction of neurological activity trends has profound implications in both medical and research contexts. This project, conducted with limited resources, leverages data collected from a Muse headset, employing LSTM and ARIMA models to navigate the challenges of predicting intricate brain activity patterns.

Project Highlights

Innovative Data Preprocessing:

Utilization of MICE imputation for missing data and Augmented Dickey-Fuller test for checking data stationarity, ensuring robust model input.

Good Performance of LSTM Models:

Preliminary results show promising RMSE (Root Mean Square Error) and MAE (Mean Absolute Error) scores for the LSTM model, indicating its potential in handling the complexity of neurological data.

Limitations and Challenges

Resource Constraints:

The project acknowledges the limitations posed by restricted access to specialized equipment and professional expertise in neurology and advanced data science.

Inconclusive Tests:

Despite achieving good performance metrics, the conclusive power of the tests remains limited due to the aforementioned resource constraints.

Insights and Observations

LSTM's Superiority:

The LSTM model shows a promising capacity to capture and predict complex patterns in time-series data, outperforming ARIMA in handling non-linear, multivariate sequences typical of neurological activity.

Resourcefulness in Data Science:

This project exemplifies how limited resources can still yield significant insights, encouraging a resourceful approach to data collection and analysis.

Disclaimer

It's important to note that this project is exploratory in nature, conducted by an enthusiast rather than medical professionals. The methodologies and findings are intended for academic and research purposes, not clinical application.