-
ในปัจจุบันยังมีการใช้สารเคมีเป็นจำนวนมากในการควบคุมและจัดการกับโรคพืช เนื่องจากประสิทธิภาพในการควบคุมโรคพืชโดยชีววิธียังไม่สามารถเทียบเคียงสารเคมีได้ หากแต่การใช้สารเคมีก็มีผลกระทบทั้งต่อสิ่งแวดล้อม ต่อตัวเกษตรกร และผู้บริโภค การได้มาซึ่งจุลินทรีย์ปฏิปักษ์ที่มีประสิทธิภาพในการควบคุมโรคพืชจึงเป็นการส่งเสริมการควบคุมโรคพืชโดยชีววิธี อีกทั้งประเทศไทยเป็นประเทศที่มีความหลากหลายทางชีวภาพเป็นอย่างมาก เป็นแหล่งอาศัยของจุลินทรีย์มากมาย คณะผู้จัดทำจึงมีวัตถุประสงค์ที่จะแยกเชื้อจุลินทรีย์จากธรรมชาติและนำมาคัดเลือกเชื้อจุลินทรีย์ที่มีฤทธิ์ปฏิปักษ์ต่อเชื้อสาเหตุโรค 3 ชนิดที่สนใจ ได้แก่ เชื้อราสาเหตุโรครากเน่าโคนเน่าของทุเรียน เชื้อราสาเหตุสาเหตุโรคใบจุดสีน้ำตาลของข้าว และเชื้อแบคทีเรียสาเหตุโรคแคงเกอร์ของพืชตระกูลส้ม
คณะเทคโนโลยีการเกษตร
Siamese fighting fish (Betta splendens) is an ornamental fish that is the first exported economically valuable fish in the country, but there is a limitation to increase the production of betta fish due to climate variability and the shortage of Thai workers. This research aims to develop 2 systems: a betta fish fry nursery system and a market-sized betta fish rearing system by using automated technology to precisely control the water quality in the system and reduce labor costs. Using precise automation consists of two systems: a minimal-waste system, which repurposes some of the waste generated from farming, and a zero-waste system, which treats and recycles all wastewater from farming. These systems aim to address issues related to water quality, animal welfare, and labor requirements in Betta fish farming. Experimental results show that these systems improve Betta fish survival rates by 10-15% compared to traditional methods. When considering net returns, the zero- waste system provides the highest profitability.
คณะวิทยาศาสตร์
This special problem aims to compare the performance of machine learning methods in time series forecasting using lagged time periods as independent variables. The lagged periods are categorized into three groups: lagged by 10 units, lagged by 15 units, and lagged by 20 units. The study employs four machine learning methods: Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). The time series data simulated as independent variables diverse including characteristics: Random Walk data, Trending data, and Non-Linear data, with sample sizes of 100, 300, 500, and 700. The research methodology involves splitting the data into 90% for training and 10% for testing. Simulations and analysis are performed using the R programming language, with 1,000 iterations conducted. The results are evaluated based on the average mean squared error (AMSE) and the average mean absolute percentage error (AMAPE) are calculated to identify the best performing method. The research findings revealed that for Random Walk data, the best performing methods are Random Forest and Support Vector Machine. For Trend data, the best performing methods are Random Forest. For Non-Linear data, the best performing methods are Support Vector Machine. When tested with real-world data, the results show that for the Euro-to-Thai Baht exchange rate, the best methods are Random Forest and Support Vector Machine. For the S&P 500 Index in USD, the best performing methods are Random Forest. For the Bank of America Corp Index in USD, the best performing methods are Support Vector Machine.
คณะวิทยาศาสตร์
Otitis Media is an infection of the middle ear that can occur in individuals of all ages. Diagnosis typically involves analyzing images taken with an otoscope by specialized physicians, which relies heavily on medical experience to expedite the process. This research introduces computer vision technology to assist in the preliminary diagnosis, aiding expert decision-making. By utilizing deep learning techniques and convolutional neural networks, specifically the YOLOv8 and Inception v3 architectures, the study aims to classify the disease and its five characteristics used by physicians: color, transparency, fluid, retraction, and perforation. Additionally, image segmentation and classification methods were employed to analyze and predict the types of Otitis Media, which are categorized into four types: Otitis Media with Effusion, Acute Otitis Media with Effusion, Perforation, and Normal. Experimental results indicate that the classification model performs moderately well in directly classifying Otitis Media, with an accuracy of 65.7%, a recall of 65.7%, and a precision of 67.6%. Moreover, the model provides the best results for classifying the perforation characteristic, with an accuracy of 91.8%, a recall of 91.8%, and a precision of 92.1%. In contrast, the classification model that incorporates image segmentation techniques achieved the best overall performance, with an mAP50-95 of 79.63%, a recall of 100%, and a precision of 99.8%. However, this model has not yet been tested for classifying the different types of Otitis Media.