SVM Textklassificering av kundmeddelanden hos ett investeringsföretag

Sammanfattning

The rapid progress of artificial intelligence (AI) and machine learning (ML) opens new avenues for streamlining customer support, notably through automated text classification. This thesis explores the potential of automating the classification of customer messages at a Swedish investment firm using Term Frequency-Inverse Document Frequency (TF-IDF) vectors and a Support Vector Machine (SVM) model. The goal is to develop a machine learning model in Swedish that can accurately identify relevant categories in customer inquiries, aiming to replace manual tagging. The data set comprised 67 684 historical customer messages distributed across 26 categories. The project includes both the technical development of the model and an organizational impact analysis from a cost-benefit and learning perspective. The results show that although the model demonstrates some classification ability (accuracy 51%, F1-score 58%), it does not yet meet the standard required for full implementation. However, the study highlights valuable organizational effects, including enhanced learning culture, more efficient workflows, and improved readiness for broader AI adoption. Thus, the thesis illustrates how even modest AI initiatives can generate long-term strategic value.