Abstract Detecting and reacting to non-technical losses (NTL) is a fundamental activity that energy providers need to face in their daily routines. This is known to be challenging since the phenomenon of NTL is multi-factored, dynamic and extremely contextual, which makes artificial intelligence (AI) and, in particular, machine learning, natural areas to bring effective and tailored solutions. If the human factor is disregarded in the process of detecting NTL, there is a high risk of performance degradation since typical problems like dataset shift and biases cannot be easily identified by an algorithm. This paper presents a case study on incorporating explainable AI (XAI) in a mature NTL detection system that has been in production in the last years both in electricity and gas. The experience shows that incorporating this capability brings interesting improvements to the initial system and especially serves as a common ground where domain experts, data scientists, and business analysts can meet.

Abstract Non-technical losses (NTL) is a problem that many utility companies try to solve, often using black-box supervised classification algorithms. In general, this approach achieves good results. However, in practice, NTL detection faces technical, economic, and transparency challenges that cannot be easily solved and which compromise the quality and fairness of the predictions. In this work, we contextualise these problems in an NTL detection system built for an international utility company. We explain how we have mitigated them by moving from classification into a regression system and introducing explanatory techniques to improve its accuracy and understanding. As we show in this work, the regression approach can be a good option to mitigate these technical problems, and can be adjusted in order to capture the most striking NTL cases. Moreover, explainable AI (through Shapley Values) allows us to both validate the correctness of the regression approach in this context beyond benchmarking, and improve the transparency of our system drastically.

The detection of Non-Technical Losses using black-box supervised classification algorithms faces technical, economic, and transparency challenges that compromise the quality and fairness of predictions. In this work, we explain how we have mitigated them in a deployed NTL detection system by moving from classification into a regression system and introducing explanatory techniques to improve its accuracy and understanding.

Predictive Business Process Monitoring is becoming an essential aid for organizations, providing online operational support of their processes. This paper tackles the fundamental problem of equipping predictive business process monitoring with explanation capabilities, so that not only the what but also the why is reported when predicting generic KPIs like remaining time, or activity execution. We use the game theory of Shapley Values to obtain robust explanations of the predictions. The approach has been implemented and tested on real-life benchmarks, showing for the first time how explanations can be given in the field of predictive business process monitoring.

Data from utility meters (gas, electricity, water) is a rich source of information for distribution companies, beyond billing. In this paper we present a supervised technique, which primarily but not only feeds on meter information, to detect meter anomalies and customer fraudulent behavior (meter tampering). Our system detects anomalous meter readings on the basis of models built using machine learning techniques on past data. Unlike most previous work, it can incrementally incorporate the result of field checks to grow the database of fraud and non-fraud patterns, therefore increasing model precision over time and potentially adapting to emerging fraud patterns. The full system has been developed with a company providing electricity and gas and already used to carry out several field checks, with large improvements in fraud detection over the previous checks which used simpler techniques.