<![CDATA[Akutusocks is a retail business that utilizes the TikTok platform for digital martaketing and as a vital two way communication channel with its consumers. The high intensity of interaction on this social media platform produces extensive review data containing various customer perceptions, complaints, and appreciation. However, this data remains largely unstructured, making it extremely difficult for the management team to analyze the information manually and efficiently. This study aims to implement advanced text mining techniques to classify the sentiment of Store Akutusocks reviews into positive and negative categories in order to provide an objective basis for evaluating the quality of product and service quality. The methodology applied in this study integrates the K-Nearest Neighbor (KNN) algorithm with a lexicon-based approach to streamline the initial data labeling process for thousands of user comments. The research stages began with rigorous text preprocessing, which is crucial for improving data quality. This process included case folding, cleansing, tokenization, and normalization to correct slang terms and abbreviations specific to TikTok, as well as stopword removal and stemming to reduce words to their base forms. Feature weighting was performed using the Term Frequency-Inverse Document Frequency (TF-IDF) method to extract dominant keywords representing user sentiment. This analysis is vital for Store Akutusocks in mitigating digital reputation risks and understanding market preferences. Through model testing using a Confusion Matrix, this study measures classification accuracy and provides deep insights into the effectiveness of the KNN algorithm]]>
Copyrights © 2026
