Abstract: Investment estimates for coal construction projects accurately and efficiently is the key to scientific decision-making for the enterprises.However, there are some problems often encountered in the early stage of project decision-making, such as lack of information, tight time, inability to complete detailed estimates, and large errors in simple calculations.Based on in-depth analysis of a large number of mine investment data, this paper proposes an investment estimation model based on least squares support vector machine(LS-SVM).By learning the implicit relationship between the technical characteristics and the investment of existing projects, determining model parameters and establishing decision functions to estimate the investment of the proposed project.According to the detailed algorithm flow, project feature extraction method, data analysis principle and process given in this paper, it is applied to the construction investment estimation of mines, and good prediction accuracy is obtained.Among them, the prediction of LS-SVM with the RBF kernel function gets the better fit.Through the analysis of the case, comparing it with the productivity index method, the LS-SVM algorithm has obvious advantages, and the prediction error can meet the allowable error requirements of the every stages of decision-making.The research results show that the investment estimation model based on LS-SVM has strong learning ability and high application value, is an effective supplement and improvement to the traditional investment estimation method, and this method can provide reliable data support for enterprise decision-making.