KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 8 of 8

Some of the metrics are blocked by your
consent settings
Stock Price Prediction from Multi Data Sources Using LSTM, FinBERT and BERTweet
(2026-06-16)
Auensupa, Suchart
;
Netisopakul, Ponrudee
;
Chotipant, Supannada
This research presents a stock price forecasting approach for technology sector companies Apple, Amazon and Tesla. The study begins by comparing a statistical model (ARIMA) with a deep learning model (LSTM) to identify the best-performing model, which is then used as the baseline for stock price forecasting. The approach integrates data from multiple sources, including numerical time series data and textual data. Numerical inputs consist of open, high, and low prices, which are used to forecast the closing price. Technical indicator features are subsequently added to enhance the model's predictive capability. Finally, textual data from economic news and Twitter social media reflecting market sentiment are incorporated. News sentiment is analyzed using the FinBERT model, while sentiment from social media data is evaluated using the BERTweet model. The resulting sentiment features are then combined with the numerical data and all inputs are processed using the LSTM model. Experimental results show that incorporating technical indicator features improves forecasting accuracy by an average of 17%. Furthermore, integrating textual data from news improves accuracy by an additional 6%, resulting in an overall performance improvement of up to 23%. These findings demonstrate the value of integrating multi-source data and highlight the important role of textual information in enhancing stock price forecasting performance.
Some of the metrics are blocked by your
consent settings
Application of Large Language Models for Aspect-Based Sentiment Analysis on Social Media Data: A Case Study of the Thai Telecommunications Industry
(2026-01-01)
Limseesawan, Krittapas
;
Netisopakul, Ponrudee
;
Chotipant, Supannada
;
Voravuthikunchai, Winn
;
Sirivorachodphokin, Thitirat
Social media is a valuable source of consumer opinion data, particularly in Thailand's highly competitive telecommunications market among AIS, TRUE, and DTAC. This study compares three LLM-based approaches for Aspect-Based Sentiment Analysis (ABSA) on 3,508 Thai-language messages from Pantip.com and YouTube.com: (1) Prompt Engineering, (2) Domain Adaptation & Fine-Tuning, and (3) Multi-Agent Debate Framework. Results show that Domain Adaptation & Fine-Tuning achieves the best accuracy-latency trade-off, with Qwen2.5-1.5B-Instruct exceeding 87% average accuracy in under 4 seconds per message. Multi-Agent Debate achieves the highest Category accuracy (82.82%) but at a latency cost of 10-16 seconds per message. Gemini-2.5-Flash-Lite provides the best speed-accuracy balance for Prompt Engineering without additional training. Critically, small open-source models (1B-1.5B parameters) subjected to domain-specific fine-tuning can approach or surpass proprietary large models on this task, suggesting domain alignment may outweigh raw parameter scale for narrow, well-defined ABSA tasks.
Some of the metrics are blocked by your
consent settings
Text Sentiment Analysis for Thailand Tourism Recommendation
(2025-01-01)
Boonyarakthunya, Ittichai
;
Chotipant, Supannada
Thailand is one of the world's most popular travel destinations. In the digital age, online reviews written by previous travelers have become a highly influential information source for decision-making. Unfortunately in many cases, users do not assign ratings, or the platforms do not require them. As a result, these systems are unable to process sentiment effectively from the textual reviews. Moreover, most existing recommendation systems lack the ability to group destinations with similar characteristics or categories. This paper proposes a tourism recommendation system that integrates sentiment analysis with tag clustering. The system is capable of processing both user sentiment and destination-related content, enabling it to generate personalized recommendations that align with users' preferences and emotional context. The experimental results show that the SVM model achieved an average sentiment classification accuracy of 94%. In contrast, tag clustering using DBSCAN presented a limitation in the form of high noise levels-over 30% of tags were classified as noise (represented by cluster -1), indicating that a significant number of tags could not be assigned to any meaningful cluster. This reflects a limitation in tag coverage and suggests that further refinement is needed to improve grouping performance in real-world datasets with diverse and contextually ambiguous descriptions.
Some of the metrics are blocked by your
consent settings
ThinkMeal: Ingredient Classification and Recipe Recommendation Application
(2024-01-01)
Aungtanagul, Napat
;
Hoontamai, Thitiwut
;
Chotiphan, Theerada
;
Chotipant, Supannada
Food is one of the four essential factors, and nowadays, Thai cuisine is highly famous domestically and internationally. The majority of Thai culinary culture involves preparing and enjoying meals within households. Therefore, it's essential to have recipes that are easily accessible and adjustable to one's taste preferences. Currently, there are applications available to assist in finding recipes. However, these applications typically limit searches to recipe or ingredient names. To address this limitation, a new application called "ThinkMeal"has been developed to recommend dishes based on available ingredients. ThinkMeal utilizes a model based on the MobileNetV2 image classification, achieving an impressive accuracy of up to 95.14% in experiments conducted with 13 types of ingredients. This enables users to conveniently find recipes that meet their preferences, either by searching recipe names and ingredient names or even by uploading images of available ingredients. This application serves as a helpful tool for individuals interested in cooking, offering convenient assistance in meal preparation.
Some of the metrics are blocked by your
consent settings
Semantic-based Thai Recipe Recommendation
(2022-01-01)
Chotipant, Supannada
Nowadays, people are constantly affected by epidemics such as COVID-19. To reduce the risk of acquiring germs in the community, people's lifestyles have been changed, and they are more inclined to cook for themselves. Typically, people can usually quickly and easily find recipe information via websites and applications. The resulting recipes consist of ingredients as specified by the user. Unfortunately, users often have ingredients that disappear in available cooking recipes. This makes the system is unable to recommend all relevant recipes to users, although the users can use the existing ingredients instead of the ingredients specified in the recipes. Based on this limitation, this research proposes a semantic-based Thai cooking recipe recommendation system which can recommend recipes based on the ingredient substitutes. This research uses existing Thai food ontology to retrieve substitute ingredients based on three different ingredient properties, such as smell, taste, and texture. To recommend cooking recipes, the system expands the given user queries with substitute ingredients and then calculates similarities between all queries and each cooking recipe. Recipes with high similarities are presented and ranked to users. To evaluate the performances, precision, recall and f-measure are applied. The experiments demonstrate that the proposed method performs well with 0.96, 0.72, and 0.82 in precision, recall, and f-measure respectively.
Some of the metrics are blocked by your
consent settings
Performance Measurement of Energy Optimal Path Finding for Waste Collection Robot Using ACO Algorithm
(2022-01-01)
Tomitagawa, Koki
;
Anuntachai, Anuntapat
;
Chotipant, Supannada
;
Wongwirat, Olarn
;
Kuchii, Shigeru
Indoor waste collection that utilizes mobile robots can solve the labor cost and manpower shortage but has the problem of limited energy resources, making it difficult to operate for long periods of time. Therefore, it is important to reduce the energy consumption for efficient waste collection. The waste collection robot can be modeled as a Capacitated Vehicle Routing Problem (CVRP), where heuristics algorithms can be deployed to search for the most energy-efficient path. This paper proposes the Ant Colony Optimization (ACO) algorithm for finding the optimal path of the waste collection robot. Energy consumption of the robot depends not only on the travel path but also on the weight of the waste it carries. Therefore, the proposed ACO algorithm utilizes the path distance and waste weight as the visibility. The travel distance and energy consumption are also used to determine the updated pheromone. Whereas the conventional and adapted ACO algorithms use only either the path distance or the waste weight as the visibility, respectively. The simulation experiments are conducted to compare the travel distance and the energy consumption that the waste collection robot takes by using the conventional, adapted, and proposed ACO algorithms. In the simulation experiments, the number of nodes, the waste weight, and the carrying capacity are used as parameters to verify the performance under the determined environment. The simulation results express that the proposed ACO algorithm provides a better energy optimal path in terms of travel distance and energy consumption than the conventional and adapted ACO algorithms.
Some of the metrics are blocked by your
consent settings
Differentiative Feature-based Fall Detection System
(2021-01-01)
Chotipant, Supannada
;
Jamsri, Pornsuree
Elderly people are dealing with falling down on a daily basis. This incident can happen anytime at any place. There is high risk of falling not only the elder but also the caregiver. Although there are numbers of applications and devices in the market for the user, the cutting-edge technology as a machine learning-based algorithm can increase effectiveness of fall detection model into device's effectiveness. The available technology is embedded accelerometer and gyroscope sensor into a smartphone provide benefit dataset. These data can be used for reducing and managing serious injury and caregiver can assist on time. The leverage performance of a Smart Steps application by including the essence of machine learning algorithm and 5-fold cross validation rises accuracy in fall detection. Thus, this paper proposed a novel method of 4 binary classification-Decision Tree, SVM, K-Nearest Neighbors, and Gradient Boosting. The focusing on acceleration magnitude, angular velocity magnitude, and difference between pre-current, current-post values are taken into account in the study. The opened dataset, MobiFall, are split into 2 groups 1) train group 80% and 2) test group 20% for gathering effectiveness result. The model's assessment measures in 4-dimension 1) accuracy, 2) precision, 3) recall and, 4) F1-Score. The results demonstrates increasing values that 95.65% of accuracy, 91.20% precision, 90.86% recall and, 91.03% F1-Score. The fall detection of the study can conclude that the machine learning-based algorithm offers more accuracy and effectively than threshold-based algorithm.
Some of the metrics are blocked by your
consent settings
A Personalized Food Recommendation Chatbot System for Diabetes Patients
(2020-01-01)
Thongyoo, Phupat
;
Anantapanya, Phuttipong
;
Jamsri, Pornsuree
;
Chotipant, Supannada
Diabetes is a disorder of the body that is unable to produce enough insulin. Diabetes causes the body to improperly burn sugar, which affects the blood sugar level leaving a sugar residue. Diabetes is related to genes, body weight, lack of exercise and aging. When patients with diabetes neglect good nutrition this can cause many health problems. This research, therefore, develops a chatbot named “Waan-Noy” to recommend a diet suitable for individuals with diabetes and build a cooperative health society. Our chatbot recommends personalized eating. It is suitable for use by diabetes patients as indicated by their evaluations. Through use of nutrition therapy controls, Waan-Noy recommends specific foods. The user’s evaluation is divided into 3 areas: content, design, and implementation to determine user degree of satisfaction with Waan-Noy.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results