Concurrent Sentiment and Concept Extraction from Corporate Annual Reports for Financial Performance Forecasting

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216275%3A25410%2F15%3A39899805" target="_blank" >RIV/00216275:25410/15:39899805 - isvavai.cz</a>

Výsledek na webu

<a href="http://www.iiis.org/CDs2015/CD2015IMC/IMCIC_2015/PapersPdf/ZA973AI.pdf" target="_blank" >http://www.iiis.org/CDs2015/CD2015IMC/IMCIC_2015/PapersPdf/ZA973AI.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Concurrent Sentiment and Concept Extraction from Corporate Annual Reports for Financial Performance Forecasting

Popis výsledku v původním jazyce

This paper aims to extract both sentiment and concept knowledge from a corpus. We extract linear combinations of features to examine the relations between sentiment categories and concepts obtained using singular value decomposition. We show that the features are related to current financial performance indicators. We use the extracted knowledge to compare decision trees algorithms (Best First Decision Tree, Alternating Decision Tree) on the one hand and Support Vector Machines and Multilayer Perceptron type Neural Networks on the other hand. The aim is to investigate the performance of document classification for a set of annual reports. We show that (1) Support Vector Machines provide the best accuracy in forecasting investment / non-investment grades, and (2) poorly performing companies frequently use commonality, human interest and negative terms and emphasize current business situation, without actively addressing problems associated with economic downturn.

Název v anglickém jazyce

Concurrent Sentiment and Concept Extraction from Corporate Annual Reports for Financial Performance Forecasting

Popis výsledku anglicky

This paper aims to extract both sentiment and concept knowledge from a corpus. We extract linear combinations of features to examine the relations between sentiment categories and concepts obtained using singular value decomposition. We show that the features are related to current financial performance indicators. We use the extracted knowledge to compare decision trees algorithms (Best First Decision Tree, Alternating Decision Tree) on the one hand and Support Vector Machines and Multilayer Perceptron type Neural Networks on the other hand. The aim is to investigate the performance of document classification for a set of annual reports. We show that (1) Support Vector Machines provide the best accuracy in forecasting investment / non-investment grades, and (2) poorly performing companies frequently use commonality, human interest and negative terms and emphasize current business situation, without actively addressing problems associated with economic downturn.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/GA13-10331S" target="_blank" >GA13-10331S: Úloha textové informace v modelech predikce finanční tísně podniků - přístupy specifické podle států a průmyslových odvětví</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2015

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

6th International Multi-Conference on Complexity, Informatics and Cybernetics, IMCIC 2015: Proceedings Vol. II

ISBN

978-1-941763-20-9

ISSN

—

e-ISSN

neuvedeno

Počet stran výsledku

6

Strana od-do

57-62

Název nakladatele

International Institute of Informatics and Systemics

Místo vydání

Winter Garden

Místo konání akce

Orlando

Datum konání akce

10. 3. 2015

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—