In the practical application of spatial load forecasting of distribution network, the available information and data are often scattered and poor. In this paper, a novel spatial load density prediction algorithm based on the Least Squares Support Vector Machine (LSSVM) is proposed to solve the problems of limited samples and difficulty in identification. In this algorithm, Grey Relational Analysis (GRA) is introduced to improve the sample selection of the Least Squares Support Vector Machine (LSSVM), and Chaos Particle Swarm Optimization (CPSO) algorithm is adopted to consummate the parameter selection of the least squares support vector machine, which improves the accuracy of spatial load density prediction of the algorithm. Based on the principle of algorithm, this paper designs a detailed implementation process for the spatial load prediction method of distribution network. The performance of this algorithm is analyzed with an example. The example calculation shows that the proposed method can effectively improve the precision of load density prediction. This work is supported by National Natural Science Foundation of China (No. 51367014).