A Serverless Intelligent Framework for Privacy-Preserving Municipal Solid Waste Classification and Circular Economy Assessment: Empirical Evidence from Deolali Pravara

Abstract

The digital transformation of municipal solid waste management in resource-constrained regions continues to face persistent challenges related to infrastructure costs, technical capacity limitations, and data confidentiality concerns. This investigation presents an integrated browser-native intelligent system designed for real-time waste categorization and generation forecasting using a zero-infrastructure computational approach. Following a comprehensive 7-phase design science methodology, the research evaluates 13 classification models and 12 forecasting algorithms using 9,839 historical records collected from the Deolali Pravara municipal jurisdiction spanning 2022-2024. Empirical findings reveal that the Random Forest ensemble method attains a classification accuracy of 93.1%, while Support Vector Regression and Linear Regression demonstrate robust predictive stability with an R² coefficient of 0.929. The framework additionally quantifies circular economy potential, estimating a material recovery value of ₹13,046,534. By executing all computational operations entirely on client-side infrastructure, this solution ensures complete data sovereignty with negligible operational expenditures, offering a scalable pathway for data-informed waste governance in developing urban contexts.