A Proposed Data Preprocessing Method for an Industrial Prediction Process

Abstract

The studied mining production chain generally divided into three principal units: Destoning, screening and loading. The role of the screening unit is to screen the phosphate produced by the destoning unit before the loading to trains. Its efficiency depends on several parameters, which makes analysis and decision making for its improvement very complicated. The purpose of this paper is to propose a prediction system to evaluate and monitor in advance the efficiency of the screening unit. This system is based on Knowledge discovery in databases that comprises generally three steps: data pre-processing, development of prediction models and finally validation and verification of the proposed models. The first consists in having in-depth information and knowledge about the application domain, in order to determine the set of parameters influencing the efficiency and to pre-process the data of the these parameters, in order to improve their quality before being used by the second step, which aims to develop predictive models that will be validated and verified with different evaluation criteria during the last step. This work focuses on the first level of development of the system in question, data pre-processing, which has been applied to the mine's screening unit facilities, and the results of this case study are also presented.