Research On Coal Mine Supplies Forecast Using Data Mining

Data mining technology belongs to a promising knowledge discovery method.By means of an appropriate data mining algorithm,the valuable information in the observation data set is extracted,and the intrinsic characteristics or relationships of the observation data are expressed in a novel and understandable method.Due to its strong decoupling ability,high feasibility,reliable results and wide application fields,the data mining technology has gradually become one of the hottest research directions in the industrial production and scientific research.Because of the prediction accuracy is affected by the quality of the prediction model,this paper mainly studies the improving method of the prediction model,and applies the improved prediction model to the supplies forecast problem of colliery.The main work and achievements are as follows:(1)The characteristics and problems of the material management system of colliery are analyzed.The related concepts and basic principles of prediction model and data mining are studied.Several common data mining methods,such as,k-means cluster,support vector machines,Bayesian methods,back propagation network,have been detailed analyzed,and the advantages and disadvantages of these methods in the data mining process are discussed in depth.(2)By studying the production process of colliery,the supplies consumption for each production part are deeply explored and the composition of colliery supplies costs is summarized.According to the characteristics of colliery,several influencing factors of colliery supplies cost are refined,such as seasonal factors,reserves factors,operational factors,and random factors.A brief analysis of these influencing factors has laid a theoretical foundation for the selection of training data and the improvement of prediction accuracy.(3)In order to overcome the defects of BP network,such as,the training process is easy to fall into local optimum;easy over-fitting or under-fitting;artificially setting iteration termination conditions can not determine the right fit of the network structure when training stops;lack of generalization ability,etc.,a colliery supplies prediction model which integrates the excellent numerical characteristics of the SSA and BP neural network is proposed.The prediction accuracy has been improved and the drawbacks of BP neural network are alleviated.The prediction experiments reveal that the SSA-BP neural network prediction model can obtain higher prediction accuracy under the considered supplies items.The prediction model inherits the nonlinear relationship expression ability and the prediction process stability of the BP neuralnetwork.The SSA is used to optimize the structure of BP neural network,and the possibility of obtaining a global optimal network structure is improved.Research findings improve the prediction accuracy of the colliery supplies prediction model,and make contributions for the formulate plans of colliery production.