Solving Old Problems with New Weapons

Proceedings, Vol. Proceedings of the 58th Conference of Metallurgists Hosting Copper 2019, 2019

Torrealba-Vargas, J.

Data mining and modelling of operational data involves the discovery of patterns in large datasets for future use in benchmarking and plant optimization. The modern pattern of discovery in large data sets involves statistical techniques, machine learning and new visualization tools. Former examples of data mining are revisited and presented in this document. INTRODUCTION A network of enthusiastic and active professionals is currently developing and maintaining several open source statistical software tools (packages) for data analysis (example Rstudio, Orange, Spyder, etc.). The advantages of the open source software philosophy are the fast development and testing of the solutions presented to the community. Books and support are available for each software application. Since each modern work activity generates massive amounts of data, it is attractive to learn and test those available tools considering that we are at the door of the new work revolution coined as “Mine 4.0”. Data mining involves collecting large datasets, the preparation and cleaning of a dataset (term named data wrangling), the uploading of the data to a database program capable of handling big datasets, the development of prediction models and the presentation of the results. Machine learning techniques help to develop mathematical models and algorithms that could be used to improve the performance on a specific task. Mine operations are producing an incredible amount of data each day and the need to analyze those results seems to be impossible due to the lack of time, as well as the lack of skills to investigate the data collected. If we add to this the traditional habit of working in silos, an additional challenge is to consolidate databases from different departments in search of trends that could help to optimize production.

Keywords: Copper 2019, COM2019