Online news for performance appraisal of Turkish Banking with Web Mining

The latest advances in communication techniques and tools shaped how organizations behave and design their competitive strategies. In this study, Web content mining is used to analyze the structure of news about Turkish banks. For this purpose, six privately owned and state owned Turkish banks, which have the highest number of news on Google Finance portal, are selected.

Keywords used in study, which are determined by an extensive literature review, are categorized as financial keywords and non-financial keywords. After information retrieval process, natural language processing techniques are employed in an order of removing whitespaces, punctuation and numbers, transforming to lowercases, tokenizing, filtering English stop words and stemming. Frequencies of keywords were extracted for each bank to inspect whether there are differences between banks with respect to news. In addition, hierarchical clustering and k-means clustering data mining techniques are applied on keyword frequencies to observe the clusters on bank news. Findings demonstrate that (i) “credit” term has a co-occurrence value of 0.84 with the “liquidity” term, (ii) the highest co-occurrence value is between “investment” and “equity” terms with a value of 0.92., (iii) the Return on Equity and expense terms are linked to financial terms with is TFIDF scores, (iv) all the other terms in dictionary has a value of 0 as TFIDF, (v) “ROE” and “expense” are two dominant terms causing differentiations between documents (bank news) in corpus and providing the most information about that specific document, (vi) the non-financial terms are demonstrated with thirteen keywords, (vii) Bank 2 has a clear superiority on frequency of term “innovation” compared to other banks, (viii) the most important terms (providing the most available information about a specific document) are “innovation”, “governance”, “strategy”, “personnel”, “employee”, “training” and “organization” in descending order.