Outlier detection in financial statements: a text mining method

Abstract

This paper presents a text mining methodology to extract outlying knowledge from a collection of financial statements. The main idea is to extract relevant financial performance indicators and discover implicit textual description of the indicators. The extracted information was represented using a network language i.e. conceptual graph. Outlier mining was performed on the conceptual graph representation using a deviation based method. Experiments were carried out to evaluate the effectiveness of the proposed method. Results show that the proposed method is able to excerpt outlying knowledge from the financial statements with accuracy comparable to human experts.

Comparative Study of Document Clustering Algorithms

“…For example, text mining had been used in biomedicine to investigate the relationship between genes and symptoms of breast cancer [1] as well as to retrieve information on the conditions of patients accurately and comprehensively through doctors" notes [2]. Meanwhile, text mining on financial statements had also been used to detect outliers in order to successfully identify fraud [3]. In social science, text mining was used on tweets from Twitter Application shared by ten libraries of renowned universities around the world to look at the similarities and differences of words used in the said social media [4].…”