# DECODE: Data-driven Energy Consumption Prediction leveraging Historical Data and Environmental Factors in Buildings

**Authors:** Aditya Mishra, Haroon R. Lone, Aayush Mishra  
**arXiv:** [2309.02908](https://arxiv.org/abs/2309.02908)  
**Date:** 2023-09-06  
**Summary page:** https://www.summarizepaper.com/en/paper/2309.02908v1/

## Key points

- Importance of energy prediction in buildings for effective energy management
- Precise predictions crucial for optimal energy consumption and distribution within the grid
- Introduction of a Long Short-Term Memory (LSTM) model for building energy consumption forecasting
- Utilization of historical energy data, occupancy patterns, and weather conditions for accurate predictions
- Comparison with established prediction methods such as linear regression, decision trees, and random forest
- LSTM model emerges as superior performer with high R2 score (0.97) and favorable mean absolute error (MAE) (0.007)
- Ability to achieve efficient forecasts even with limited dataset
- Addressing concerns about overfitting and underfitting through rigorous training and evaluation on real-world data
- LSTM model outperforms alternative methods in terms of efficiency, generalizability, and reliability

## Summary

The paper titled "DECODE: Data-driven Energy Consumption Prediction leveraging Historical Data and Environmental Factors in Buildings" by Aditya Mishra, Haroon R. Lone, and Aayush Mishra focuses on the importance of energy prediction in buildings for effective energy management. The authors highlight that precise predictions are crucial for achieving optimal energy consumption and distribution within the grid. To address this need, the paper introduces a Long Short-Term Memory (LSTM) model specifically designed to forecast building energy consumption. The LSTM model utilizes historical energy data, occupancy patterns, and weather conditions to provide accurate short, medium, and long-term energy predictions for both residential and commercial buildings. In order to evaluate the performance of their LSTM model, the authors compare it with established prediction methods such as linear regression, decision trees, and random forest. Encouragingly, the proposed LSTM model emerges as the superior performer across all metrics. It demonstrates exceptional prediction accuracy with a high R2 score of 0.97 and a favorable mean absolute error (MAE) of 0.007. One notable advantage of the developed LSTM model is its ability to achieve efficient energy consumption forecasts even when trained on a limited dataset. The authors address concerns about overfitting (variance) and underfitting (bias) through rigorous training and evaluation on real-world data. Overall, this research significantly contributes to energy prediction by offering a robust LSTM model that outperforms alternative methods in terms of efficiency, generalizability, and reliability. The findings highlight the potential of utilizing data-driven approaches like LSTM models for accurate energy consumption forecasting in buildings.