Data Collection and Pre‐processing

Summary

Artificial Intelligence (AI) model performance is dependent on the quality, quantity, and variety in training data. While low‐quality data, which is defined by errors, incompleteness, inconsistencies, or biases, can result in less‐than‐ideal model performance and poor decisions, high‐quality data guarantees accurate forecasts, trustworthy results, and trust. This chapter explores data collecting, pre‐processing, and annotation strategies as they relate to AI applications for power electronics. The chapter also discusses ways to maintain ethical compliance and data privacy, especially in power electronics applications where operational data needs to be managed carefully. The chapter contains a case study that illustrates data smoothing and outlier detection using MATLAB in order to offer useful insights.