A Next-Generation Hybrid TCN–Transformer Encoder Deep Learning Framework for Robust and Interpretable Climate Vulnerability Index Prediction in India

View at publisher Summarise with VIRAT
Authors: Jakka Vidyadhari, V. Sathyendra Kumar
Cited by: 0
References: 1

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Climate change continues to intensify socioeconomic risks across India, making accurate prediction of the Climate Vulnerability Index (CVI) essential for proactive planning and resilience building. Existing deep learning approaches such as LSTM, Bi-LSTM, GRU, Transformers, and Autoencoders capture temporal patterns but often struggle to model long-range dependencies and multi-indicator interactions simultaneously. To address these limitations, this paper proposes a next-generation hybrid deep learning framework that combines a Temporal Convolutional Network (TCN) with a Transformer Encoder for high-precision CVI forecasting using multivariate climate, environmental, and socio-economic indicators from the NITI Aayog NDAP platform. The TCN component effectively captures temporal hierarchies and long-distance patterns through dilated causal convolutions, while the Transformer Encoder leverages multi-head self-attention to learn complex cross-indicator relationships and interpret feature contributions. Experimental results demonstrate that the proposed hybrid model significantly outperforms traditional recurrent architectures, achieving an RMSE of 0.052, an MAE of 0.031, and an R2score of 0.47, highlighting superior stability and predictive power. The framework provides decision makers with a reliable tool to identify emerging vulnerability hotspots and design targeted climate adaptation strategies across India.

Fields of Study & Subjects

Climate Change and Environmental Impact Stock Market Forecasting Methods Disaster Management and Resilience

Publication details

Journal / source
International Conference on Computing, Communication, Control and Cyber-Physical Systems (I5CPS)
Pages
1–4
Publication date
26 February 2026
Language
EN