Automated Email Sorting in Organizations: A Machine Learning based Multi-Class Classification System

Email communication enables seamless communication, collaboration, and decision-making processes across diverse organizational structures. The ability to efficiently organize, analyze, and retrieve insights from emails has become increasingly important. This paper addresses the challenge of automating email classification into categories such as Sales, HR, and Development based on their content. The dataset used in this study combines various open-source email datasets, including Enron and Atari. The analysis utilized sophisticated feature extraction techniques, such as TF-IDF and Word2vec, alongside systematic hyperparameter optimization using 5-fold cross-validation, ensuring robust model performance. We applied four classification algorithms— Linear Support Vector Classifier, Random Forest Classifier, Multinomial Naïve Bayes and Logistic Regression to perform the classification task. Experimental results demonstrate a peak accuracy of 96% from proposed model of combined feature vectors. The proposed models and methodologies are applicable for classifying English-language texts and pave the way for future research in this field.