Text analysis and knowledge mining system

, and the workshop participants at the University of Michigan for their comments. All errors remain mine. AbstractI test the stock market efficiency with respect to the information in the texts of annual reports. More specifically, I examine the implications of corporate annual reports' risk sentiment for future earnings and stock returns. I measure the risk sentiment of annual reports by counting the frequency of words related to risk or uncertainty in the 10-K filings. I find that an increase in risk sentiment is associated with lower future earnings: Firms with a larger increase in risk sentiment have more negative earnings changes in the next year. Risk sentiment of annual reports can predict future returns in a cross-sectional setting: Firms with a large increase in risk sentiment experience significantly negative returns relative to those firms with little increase in risk sentiment in the twelve months after the annual report filing date. A hedge portfolio based on buying firms with a minor increase in risk sentiment of annual reports and shorting firms with a large increase in risk sentiment generates an annual Alpha of more than 10% measured using the four-factor model including the Fama-French three factors and the momentum factor.

show abstract

“…1 In today's world, a huge amount of information is stored as text instead of numeric data (Nasukawa and Nagano (2001)). …”

Section: Introductionmentioning

confidence: 99%

Do Stock Market Investors Understand the Risk Sentiment of Corporate Annual Reports?

Li¹

2006

SSRN Journal

235

165

View full text Add to dashboard Cite

show abstract

“…Some text visualization applications focus on visualizing text mining results and showing relationships among terms, as well as text patterns in document collections [22] [23]. Other techniques such as tag cloud [24] and WordTree [25] focus on providing visual summaries of the content and relations of text corpus.…”

Section: Text Visualizationmentioning

confidence: 99%

Analysis of large digital collections with interactive visualization

Esteva

Jain

et al. 2011

2011 IEEE Conference on Visual Analytics Science and Technology (VAST)

View full text Add to dashboard Cite

To make decisions about the long-term preservation and access of large digital collections, archivists gather information such as the collections' contents, their organizational structure, and their file format composition. To date, the process of analyzing a collection -from data gathering to exploratory analysis and final conclusions -has largely been conducted using pen and paper methods. To help archivists analyze large-scale digital collections for archival purposes, we developed an interactive visual analytics application. The application narrows down different kinds of information about the collection, and presents them as meaningful data views. Multiple views and analysis features can be linked or unlinked on demand to enable researchers to compare and contrast different analyses, and to identify trends. We describe and present two user scenarios to show how the application allowed archivists to learn about a collection with accuracy, facilitated decisionmaking, and helped them arrive at conclusions.

show abstract

“…This knowledge discovery process has many domains and application areas such as in bioinformatics [2], business analytics [3], text analysis [4], web data analysis [5], health care [1] and many other domains where there is scope for hidden information retrieval.…”

Section: Introductionmentioning

confidence: 99%

A Parallel Fuzzy-Genetic Algorithm for Classification and Prediction

Abounaser¹,

Talkhan²,

Fahmy³

2016

ijacsa

View full text Add to dashboard Cite

Abstract-One of the top challenging problems in data mining domain is the distributed data mining (DDM) and mining multiagent data. In distributed environment, classical techniques require that the distributed data be first collected in a data warehouse which is usually either ineffective or infeasible. Hence, mining over decentralized data sources can overcome such issues. Rule-based classifiers involve sharp cutoffs for continuous attributes. Fuzzy Logic System (FLS) has features that make it an adequate tool for addressing this shortcoming effectively and efficiently. In this paper, a framework for a Parallel FuzzyGenetic Algorithm (PFGA) has been developed for classification and prediction over decentralized data sources. The model parameters are evolved using two nested genetic algorithms (GAs). The outer GA evolves the fuzzy sets whereas the inner GA evolves the fuzzy rules. During optimization, best rules are only distributed among agents to construct the overall optimized model. Several experiments have been conducted over many benchmark datasets. The experiment results show that the developed model has good accuracy and more efficient in performance and comprehensibility of linguistic rules compared to some models implemented in KEEL software tool.

show abstract

Text analysis and knowledge mining system

Cited by 154 publications

References 9 publications

Do Stock Market Investors Understand the Risk Sentiment of Corporate Annual Reports?

Do Stock Market Investors Understand the Risk Sentiment of Corporate Annual Reports?

Analysis of large digital collections with interactive visualization

A Parallel Fuzzy-Genetic Algorithm for Classification and Prediction

Contact Info

Product

Resources

About