Reporting clustering and creating demand-based report recommendations with natural language processing (NLP) in financial industry

With the rapid development of technologies in the field of artificial intelligence, the interests of both academia and various sectors have been turned to this field. In this context, natural language processing and machine learning studies, which are among the artificial intelligence technologies, have increased considerably in sectors with a wide variety of customers and products. Natural language processing, document classification - clustering, named entity recognition, sentiment analysis, spell checker, chatbots, language translation, etc. has study subjects. In this study, a deep learning-based natural language processing and machine learning study was carried out by using the report request contents of a private bank in the financial sector. Within the scope of the study, the representations of the report request contents were created with TF-IDF, Word2vec and Doc2vec. K-Means, K-Medoids and Agglomerative machine learning clustering algorithms were clustered and report requests in the same cluster were listed. In the study, clustering successes were interpreted with Silhouette coefficient, Calinski-Harabasz and Davies-Bouldin indices. With the study, if there is a report request made similar to a new report request, it is aimed to list this report request or report requests. As a result of the study, the best clustering result was obtained with K-Means in all word embedding methods. In the clustering with Word2vec and Doc2vec, it was seen that the evaluation metrics gave similar results, and in the similarity study, the results obtained with the three word embeddings method were similar.