ePrints@IIScePrints@IISc Home | About | Browse | Latest Additions | Advanced Search | Contact | Help

A Hybrid Deep Learning Model to Predict Business Closure from Reviews and User Attributes Using Sentiment Aligned Topic Model

Thazhackal, SS and Devi, VS (2019) A Hybrid Deep Learning Model to Predict Business Closure from Reviews and User Attributes Using Sentiment Aligned Topic Model. In: Proceedings of the 2018 IEEE Symposium Series on Computational Intelligence,, pp. 397-404.

[img] PDF
Ieee_Sym_Ser_Com_Int_397_2019.pdf - Published Version
Restricted to Registered users only

Download (447kB) | Request a copy
Official URL: https://dx.doi.org/10.1109/SSCI.2018.8628823

Abstract

Business closure is a very good indicator for success or failure of a business. This will help investors and banks as to whether to invest or lend to a particular business for future growth and benefits. Traditional machine learning techniques require extensive manual feature engineering and still do not perform satisfactorily due to significant class imbalance problem and little difference in the attributes for open and closed businesses. We have used historical data besides taking care of the class imbalance problem. Transfer learning also has been used to tackle the issue of having small categorical datasets. A hybrid deep learning model has been proposed to predict whether a business would be shut down within a specific period of time. Sentiment Aligned Topic Model (SATM) is used to extract aspect-wise sentiment scores from user reviews. Our results show a marked improvement over traditional machine learning techniques. It also shows how the aspect-wise sentiment scores corresponding to each business, computed using SATM, help to give better results. © 2018 IEEE.

Item Type: Conference Paper
Additional Information: Copyright for this article belongs to Institute of Electrical and Electronics Engineers Inc.
Keywords: Data mining; Forecasting; Learning algorithms; Machine learning, Categorical datasets; Class imbalance problems; Feature engineerings; Hybrid neural networks; lexicon generation; Machine learning techniques; Topic Modeling; yelp, Deep learning
Department/Centre: Division of Electrical Sciences > Computer Science & Automation
Depositing User: Id for Latest eprints
Date Deposited: 15 Apr 2019 05:18
Last Modified: 15 Apr 2019 05:18
URI: http://eprints.iisc.ac.in/id/eprint/62091

Actions (login required)

View Item View Item