LSTM and Simulated Annealing for Entrepreneurship Path Planning in Smart Rural Contexts

Abstract

Smart villages are evolving from “digital access” to “data empowerment” and “value co-creation”, yet challenges like infrastructure imbalance, information silos, and talent loss hinder sustainable development. College students' skill entrepreneurship offers a promising way to enhance rural human capital and industrial resilience, but is often limited by demand uncertainty and resource mismatch. This paper integrates the Long Short-Term Memory (LSTM) network with the Simulated Annealing (SA) algorithm to develop a joint framework for demand forecasting and entrepreneurial path optimization. Using multi-source data over three years, the model predicts fine-grained resource needs and searches for optimal action sequences. Experimental results show that the framework improves prediction accuracy and decision-making efficiency. A multi-agent coordination mechanism is also proposed to address privacy and energy concerns. The study provides a data-driven approach to rural revitalization and supports high-level talent engagement in grassroots entrepreneurship.