Developing an Email Classification algorithm.

Ali Khalil, Dipl. Ing.

Diplomová práce

Abstract:

Artificial intelligence and machine learning have shown the ability to automate decision-making and improve different aspects of human lives. This thesis uses the power of machine learning to build classification models that can be used to recognize the topic of an email based on its context. The purpose is to have a model that can be deployed in companies and used to correctly detect the topic or category of each email and help the employees to prioritize their email reading based on the topic. Another application may be to direct an email to the correct/suitable person or department depending on its topic. Bearing the above in mind, this thesis is composed of two main parts. The first part is a theoretical part where an overview of the basics of artificial intelligence, machine learning, natural language processing, and artificial neural networks is presented. The second part is a practical part that shows the practical steps followed in order to build the email classification model. The practical part itself includes two sequential steps. The first step is data preprocessing within which the emails are cleaned and converted into a numerical form making them suitable for the training of machine learning mathematical models. The second step is, of course, the training and testing of the machine learning models. For comparison purposes, four machine learning models have been trained. Those models are Logistic Regression, Random Forest, and Naive Bayes in addition to Artificial Neural Networks. The results confirm the applicability of the email’s classification concept and also affirm the linear nature of the classification problem in scope based on the high prediction accuracy reached by the Linear Logistic Regression. As a result of the performed analysis, the Logistic Regression and Artificial Neural Networks have shown the highest performance among the four compared algorithms.

Keywords

Artificial Intelligence, Machine Learning Algorithms, Artificial Neural Network, Natural Language Processing, Activation Functions

Jazyk práce: angličtina

Datum vytvoření / odevzdání či podání práce: 29. 11. 2022

Obhajoba závěrečné práce

• Vedoucí: doc. Ing. Arnošt Veselý, CSc.

Česká zemědělská univerzita v Praze

Provozně ekonomická fakulta

Magisterský studijní program / obor:
Systems Engineering and Informatics / Informatics