IMPLEMENTASI NATURAL LANGUAGE PROCESSING PADA APLIKASI CHATBOT BERBASIS ARTIFICIAL INTELLIGENCE UNTUK LAYANAN INFORMASI PENAWARAN HARGA (STUDI KASUS: BRANZ BSD)

Abstract

The research objective is to examine the impact of liquidity, company size, capital structure, and sales growth on company value in the transportation industry listed on the Indonesia Stock Exchange (IDX) from 2018 to 2021. Researchers used a sample of companies in the transportation sector listed on the IDX in 2018 to 2021. In the study, the population used consisted of 48 companies. The researcher applied a purposive sampling technique to select the research sample. The selection process was based on predefined criteria, resulting in 9 companies and a total of 36 financial reports. The research data consists of secondary data which is financial report data and annual reports. This study uses a data analysis approach that involves the use of descriptive statistics, followed by an examination of classical assumptions. These assumptions include several tests, including tests for normality, multicollinearity, autocorrelation, and heteroscedasticity. F-testing, t-testing, and testing the coefficient of determination are used in testing hypotheses. The research findings indicate that firm value is simultaneously influenced by liquidity, firm size, capital structure, and sales growth. Firm value has a positive influence on liquidity, capital structure, and sales growth partially. Meanwhile, firm value has a partial negative effect on firm size. The growing competitiveness of the property sector has increased the importance of providing responsive and accurate quotation price information to prospective buyers. The limited use information and communication technology causes the communication process to be less efficient, resulting in difficulties for potential buyers in obtaining the information they need. This study aims to improve the efficiency of delivering price offer information, implement a chatbot that is responsive to prospective buyers inquiries, and enhance the accuracy of responses through the application of Natural Language Processing. The research methodology comprises data collection through observation and interviews, a literature review, preprocessing data, model development using the Naïve Bayes algorithm, system implementation and evaluation. Preprocessing data includes cleaning, case folding, stopword removal, tokenization, and stemming. The result demonstrate that the application of Natural Language Processing with the Naïve Bayes method significantly improves the chatbot ability to interpret user queries and produce relevant responses. The Confusion Matrix based evaluation indicates that the system exhibits outstanding performance, achieving an accuracy of  99.7%, precision of 100%, recall of 99%, and F1-Score of 99.5%, with results supporting it is use for information dissemination.