Early childhood plays an important role in forming the basis of development, which involves stimulation of various aspects such as moral religious values, social emotional, language, cognitive, and physical motor skills. The concept of early childhood learning is focused on play, where every activity is designed to be play, so that learning becomes more effective. Parents also need to understand today's children's education to interact with children positively. This research focuses on sentiment analysis of children's education-based app reviews on the Google Play Store, using Random Forest and Logistic Regression methods. The review data is taken from three apps with the theme of child development, namely "About Kids", "PrimaKu", and "Teman Bumil", with a range of review years between 2018 and 2023. The test results show that Logistic Regression has higher accuracy compared to Random Forest, especially in the "About Kids" and "PrimaKu" applications with accuracy above 90%. The conclusion of this research highlights the importance of sentiment analysis in improving understanding of user responses to children's education applications, with suggestions for future research to increase the number of datasets and variations in testing schemes by tuning hyperparameters to improve prediction accuracy and more optimal results.

In the current era of technological advancement, deep learning has become a widely discussed and utilized topic, particularly in image classification, object detection, and natural language processing. A significant development in deep learning is the Convolutional Neural Network (CNN), which is enhanced with various optimizations such as Adam, RMSProp, and SGD. This thesis implements the Inception v3 architecture for the deep learning model, utilizing these three optimization methods to classify malaria disease. The study aims to evaluate performance and determine the best optimization based on classification accuracy. The results indicate that the SGD optimization with a learning rate of 0.001 achieved an accuracy of 94%, RMSProp with learning rates of 0.001 and 0.0001 achieved an accuracy of 96%, and Adam with learning rates of 0.001 and 0.0001 achieved an accuracy of 95%.

Chronic diseases in humans are very difficult to detect visually, for example glaucoma, hypertension, diabetes, and others. So it takes a lot of time for further medical examination by visiting a health center or hospital. Therefore, this research aims to find a solution combining medical and computer science to classify quickly and precisely. Classifying eye images requires good features and characteristics so that disease images can be classified. This research uses the Deep Learning method, namely Convolutional Neural Network with MobileNet-V3 architecture which can extract features from large resolution images very well. This research resulted in accurate classification of images of chronic diseases Normal, Diabetes, Glucoma, Cataract, Age related macular degeneration, Hypertension, Pathalogical Myopia. uses the MobileNet-V3 architecture, with transfer learning reaching 81%, and loss only 0.4913.

Rohingya in Indonesia has become trending conversation on social media. Sentiment analysis can get public responds. Big data makes the problem time efficiency labeling process, therefore the lexicon dictionary is needed for the labeling process. Data is growing and circulating very rapidly so it takes a fast and efficient time. Although it is fast and makes it easier to solve problems, it is still necessary to question the accuracy produced when using the lexicon labeling. A comparison of the labeling process between the InSet lexicon and the VADER lexicon was conducted to determine the accuracy of the labeling. It was done by combining lexicon with machine learning method of support vector machine and TF-IDF weighting and accuracy result calculated using confusion marix. Data from social media X as many as 9117 lines and labeled with InSet lexicon result 5241 negative sentiments, 1369 positive, and 521 neutral. Then the labeling results with VADER produced 2749 positive, 2523 negative, and 1881 neutral. After labeled, processed SVM and calculated accuracy with results of InSet lexicon accuracy having an average of 85.8% while the VADER SVM lexicon has an average of 82.65%.