From: Comprehensive review of text-mining applications in finance
Study | Datasets | Techniques/algorithms used | Evaluation parameters | Performance (on the basis of evaluation parameters) | References |
---|---|---|---|---|---|
Identifying text patterns for financial performance | Annual reports of US-listed companies (10-K) | Clustering, sentiment analysis | Correlations between text patterns in company’s reports and its sales performance | – | Lee et al. (2018) |
Company sustainability report analysis | Corporate disclosures | n-grams, content analysis | Variations in disclosure | − 8.89 to 1.57% | Aureli (2017) |
Competitive analysis from social media | Social media data from Social Mention, SABI database | Social mention tool | Pearson correlation, F-ratio | F-ratio: 3.361 (Good fit) | Gemar and Jiménez-Quintero (2015) |
Automatic classification of accounting literature | Articles from EbscoHost online academic database | Bayes classifier, decision tree, rule-based classifier | Accuracy | 87.27% | Chakraborty et al. (2014) |
Financial reports analysis | Quarterly reports of telecom manufacturers | Prototype-matching text clustering | Change in tone of financial reports with respect to the company’s performance | – | Kloptchenko et al. (2004) |
Computer-aided analysis of corporate disclosures | Annual reports of publicly listed DAX companies | Dictionary model, topic extraction, link analysis | Frequency of keywords, trend analysis | – | Matthies and Coners (2015) |
Deviation detection in financial statements | Financial statements | Link Grammar Parser, conceptual graphs | Cumulative similarity | – | Kamaruddin et al. (2007) |
Corporate bankruptcy prediction | Japanese annual reports | Conditional probability, chi-square test | Word frequency statistical metrics | – | Shirata et al. (2011) |
Financial statement fraud detection | Financial statements | Bag-of-words, SVM | Accuracy, recall, precision, purity, F-measure | Conceptual paper; hence, performance not evaluated | Gupta and Gill (2012) |
Financial footnote analysis | Income tax footnotes from financial reports | NB, k-means, KNN, SVM, decision tree | Runtime, accuracy, RMSE, absolute error | NB performed best Runtime: 4Â s Accuracy: 82.86% Absolute error: 0.171 RMSE: 0.414 | Heidari and Felden (2015) |