Analysis of Financial Transactions using Machine Learning

Abstract: Many people want to know the socio-ecological impact of the goods they purchase. In this thesis, we describe a system that computes the socio-ecological impact of those goods by analyzing uncategorized financial transactions. The computation is made possible by extending a system that can computate socio-ecological impact from categorized transactions. The extension further includes visualizations on the system’s web GUI using AngularJS and extension of the system’s Node.js API. To compute the socio-ecological impact the report describes a categorization service. To connect the service to the core system a RabbitMQ message queue was used. The service trained supervised machine learning models using Apache Spark’s machine learning library (MLlib) on a dataset containing about 2.4 million categorized transactions. This achieved a categorization accuracy of 82.9%. The main focus for future work is to increase accuracy by using named-entity recognition and splitting up the categorization into two steps using multiple categorizers.