Masoumeh Rostam Niakan Kalhori - Assistant Professor, Department of Energy Economics, Niroo Research Institute, Iran
Iman Taheri Emami - Ph.D. Candidate, Department of Electrical Engineering, Amirkabir University of Technology, Iran
Masoud Hasani Marzooni - Assistant Professor, Department of Energy Economics, Niroo research institute, Iran

   In this paper, predictive data mining models are employed to get insights into the efficiency of a deregulated electricity market. The bidding data of Iranian generation units in a two-phase approach are classified. Firstly, common factors that could contribute to investigating the efficiency of generation units’ bidding behavior are identified by feature selection algorithms. Then, classification rule mining algorithms are applied to extract if-then rules related to bidding blocks of generation units. The three most-applicable algorithms for classification rule mining are compared statistically. The two first algorithms are decision trees based on a direct approach. Finally, the third algorithm is the sequential covering method, perceived as an indirect approach to classification rule mining. The extracted rules are of significant importance for wholesale electricity market monitoring units (MMUs) to evaluate the market and its players thoroughly. The experimental results indicate that the partial decision tree outperforms other investigated methods.

Deregulated electricity markets, market monitoring, Data mining, Machine Learning, classification rule mining