In electric energy sales, pricing benefits those who buy a fix quantity of energy. Therefore, predictions are needed if someone wants to make a significant profit.
In the first part of this thesis, I review the problem to be solved, and the tools I used to solve it. I give a short explanation why temperature data can be useful for predicting electric energy consumption, the fact on which my models are based on. I elaborate upon data mining, mainly upon the steps of CRISP-DM, which I followed during my work. I review RapidMiner, the program I used, its developement surrounding and its operators. I explain the mathematical background of the models I later used.
In the second part, I give full details of the work I made during the semester. The documentation is built around the phases of CRISP-DM. This starts with the understanding of the task itself and the given data sets. Then comes the preparation made on these data sets to get it ready for modeling. I execute the earlier explained algorithms with many different settings and add methods that may help getting better results. I include those results in tables and compare the models through them. Then, I exemplify the best model on examples.