Research on tobacco market grid division model based on multisource data fusion and machine learning algorithms

In order to solve the problems of insufficient accuracy and insufficient consideration of spatial heterogeneity in the traditional tobacco market mesh generation method, this study proposes an improved model combining multi-source data and machine learning algorithm. First, build a distributed data processing platform based on Hadoop/spark, integrate and clean multi-source heterogeneous data such as population, geography and consumer behavior, and realize distributed storage and management through HBase; Secondly, the improved k-means++clustering algorithm is used for initial mesh generation, and the clustering process is optimized by introducing Voronoi diagram spatial constraints and contour coefficient; Then, the stochastic forest and xgboost algorithm are integrated to build a market demand forecasting model, which realizes the dynamic adjustment and optimization of the initial grid. This study provides a data-driven solution for the refined management and optimal allocation of resources in the tobacco market.