Intergration of fine-grained sentiment analysis and deep neural network to analyze financial news | |
| Author | Chawisa Phumdontree |
| Call Number | AIT Thesis no.IM-18-06 |
| Subject(s) | Artificial Intelligent Machine Learning |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Information Management, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. IM-18-06 |
| Abstract | Due to the current technological advancement of the denominated Artificial Intelligent (AI) and Machine Learning trends along with the undeniable of the big data and FinTech influencers, the novel SentiFine framework was introduced for facilitating the financial analyst or specialist who has to understand the circumstance of the financial and economic from the daily news. The objective of this framework is to analyze the sentiment in the Thai financial daily news by integrating the fine-grained sentiment analysis technique with the deep neural network. Based on the proposed SentiFine framework which included the conceptual, processes and procedures, the prototype of the SentiFine web-based system with the sufficient features was developed. The development of the minimum viable SentiFine web-based system has been built by using four main phases: system analysis, system design, system implementation, and system testing. The requirement determination and analysis were scrutinized under the system analysis phase for supplying the information to outline the physical, logical, and architecture system in the design phase. In the system implementation phase, three main tasks have been distributed: Environment Preparation, Data Model Formulation, and Web-based System Development. The concept of Thai Natural Language Processing (Thai NLP) was employed in the feature engineering part which applied only the unstructured text along with the adoption of the proposal of the United CNN Bidirectional Gated Recurrent Unit neural network model (UCBGRU) which is the integration between Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (Bi-GRU). Two experiments have been conducted for testing the accuracy and usability of the system from 12 potential participants. The accuracy of the sentiment by the user is rather low than the model evaluation part because of the insufficient amount of our dataset and the source of the ground truth, the unique characteristics of Thai NLP, and the step in data pre-processing process. For the usability test, the participants expressed their feeling on with the relatively high rating on the effectiveness, efficiency, and satisfaction factors. The outcome of this study is a prototype of SentiFine web-based system that follows with the SentiFine framework, a novel UCBGRU model, and the Thai Financial news dataset |
| Year | 2018 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis ; no. IM-18-06 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Information Management (IM) |
| Chairperson(s) | Vatcharaporn Esichaikul; |
| Examination Committee(s) | Dailey, Matthew N.;Chutiporn Anutariya; |
| Scholarship Donor(s) | Royal Thai Government Fellowship; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2018 |