<![CDATA[The questionnaire is a data collection media done by giving written questions. Some agencies this day still have not used digital media as a medium. The use of digital media technology for filling out questionnaires can ease employees' work in the process of sentiment analysis, where data is automatically stored in the database and generates sentiment analysis results filling out the questionnaire. Based on this problem, we need systems that can automatically perform sentiment classification to make it easier for employees to get the report results used as an evaluation of the Local Government Clinic. The classification executed through several stages starting from pre-processing which consists of cleaning, case folding, tokenization, filtering, and stemming. Then proceed with the calculation of term weighting and cosine similarity to the sentiment classification process using Improved K-Nearest Neighbor method. Based on the results of the implementation and testing of the system using k-fold cross-validation get an average Precision value of 0.7657, a Recall value of 0.8088, F-Measure value of 0.7847, and accuracy value of 0.7738. In testing the amount of training data, the best test results were obtained with an average Precision value of 0.8544, Recall value of 0.78, F-Measure value of 0.8151, and an accuracy value of 0.823. It takes accuracy in determining the documents used as training data on the system to produce optimal accuracy. Some words in the document still use non-standard words, causing the system to be wrong in classifying sentiments. It would be nice if the documents used training data were normalized first to obtain optimal results.]]>
Copyrights © 2022
