Thanlyin Technological University

In outcome base education, we need to utilize students' feedback as one of the key inputs to determine the effectiveness of teaching and learning activities of a university. Analyzing only rating scores does not reveal the true sentiment of students. Textual comments are more reflected true sentiment of the students. This paper is to implement a feedback analysis system based on rating score and textual comments. We used two types of dataset which include rating scores and textual comments, respectively. K-means clustering algorithm is used to cluster rating scores. Classification models are built with various classification algorithms using labeled dataset that got from clustering step. For textual comment analysis, we used Naïve Bayes classifier to train a model and classify test dataset into negative and positive sentiments using 10-fold cross validation. We also tested the model to classify the unlabeled textual comments. According to experimental results, we found that Support Vector Machine gives the best precision result for rating score classification. For textual comment analysis, Naïve Bayes classifier gives optimum result.

The procedures in the world of welding all look simple, but all welding that looks simple must have extensive knowledge in accordance with the definition of welding. This welding is a permanent connection technique, when the welding process takes place the results do not always meet the quality standards that have been determined. The cause of the welding results that do not meet the quality standards in the form of dirt on the material when before welding and so forth. The purpose of this study is that the specimen specimens are accepted according to the welding institute's standard acceptance criteria (TWI) This research uses st 37 steel by using SMAW welding electrodes LB 52U (E 7016) diameter 2.6mm for root, and LB 52U (E 7016) diameter 3.2 for fillers and caping with welding position 1G. Testing of welding results is carried out by the non destructive test of the dye penetrant method whose inspection results are determined based on the standard ASME. The results of this test study discontinuity in the form of porosity on the surface of the welding results in specimen 1, namely material contaminated by dirt, there are porosity defects scattered with a length of 6mm and porosity of 10mm so based on the standard TWI both test specimens were declared rejected, and in specimen 2, material that was not contaminated with impurities, there were porosity defects of 2mm and 2.5mm so that according to the standards of the two specimens the test was accepted. For the results of welding defects that are rejected can still be corrected by linking to the defects that, and can be done welding again.

Disturbances of turbidity and low illuminance are problems in real sea areas when recognizing objects with cameras. Therefore, the recognition target was made to emit light so that it can be recognized correctly even in that environment. However, a suitable light intensity of the target was not decided and it is obvious that recognition results was changed by light intensity of the target. This paper presents the analysis of recognition accuracy of the Real-time 3D estimation system by changing the current value of each color LED (red, green, blue) under turbid and low illuminance. Recognition experiments were conducted at the distance 600 [mm] between the ROV and 3D marker. The turbidity level was set constant value. The current value was changing from 0 [mA] to 16 [mA] for each LED individually. The best current for each LED was optimized by the fitness value and estimation value of position and orientation. The results showed that the recognition accuracy of the proposed system was improved by using optimized lighting intensity.