No abstract

Instagram is a social media that has the potential to be used to increase awareness of a product. Approximately 70% of users spend their time searching for a product on Instagram. Many people promote their products with a lack of attention to the target. So that not infrequently the information distributed is inaccurate information and not following user characteristics. This study aims to cluster the characteristics of Instagram users based on hashtag compatibility. The method used in this study is the K-Means Clustering method. Based on the results of the experiment, this research succeeded in clustering Instagram users based on the hashtag match on the text caption. Besides, TF-IDF can be used as a feature suitable for the K-Means Klastering method. The results of the hashtag "#kopi" analysis resulted in hashtag suggestions that can be used for the promotion of a product related to coffee, including the hashtag #coffeeshop and #coffee with total usage of 14968 captions.

One of the problems related to journal publishing is the process of categorizing entry into journals according to the field of science. A large number of journal documents included in a journal editorial makes it difficult to categorize so that the process of plotting to reviewers requires a long process. The review process in a journal must be done planning according to the expertise of the reviewer, to produce a quality journal. This study aims to create a classification model that can classify journals automatically using the Cosine Similarity algorithm and Support Vector Machine in the classification process and using the TF-IDF weighting method. The object of this research is abstract in scientific journals. The journals will be classified according to the reviewer's field of expertise. Based on the experimental results, the Support Vector Machine method produces better performance accuracy than the Cosine Similarity method. The results of the calculation of the value of precision, recall, and f-score are known that the Support Vector Machine method produces better amounts, in line with the accuracy value.

Classification of text with a large amount is needed to extract the information contained in it. Student comments containing suggestions and criticisms about the lecturer and the lecture process on the learning evaluation system are not well classified, resulting in a difficult assessment process. So from that, we need a classification model that can classify comments automatically into classification categories. The method used is the Cosine Similarity method, which is a method for calculating similarities between two objects expressed in two vectors. The data used in this study were 1,630 comment data with several different categories. The test in this study uses k-fold cross-validation with k = 10. The results showed that the percentage accuracy of the classification model was 80.87%.

The promotion of goods or services is now facilitated by the dissemination of information through Instagram. Dissemination of information is usually done by influencers or promotional accounts. The account used certainly has a lot of followers. Because of the large amount of follower data in that account, it can be grouped into the same characters. This is done to determine the potential for promotion using social media accounts. This study uses data from 2 popular accounts. The first account is an artist with the username ayutingting92. The second account is Infounjaya, the official promotion account from Jenderal Achmad Yani University, Yogyakarta. The results of grouping can divide follower data into two cluster groups with different interactions. The basic difference between the two groups is the number of likes and comments. The infounjaya account analysis results showed that of 4,906 followers, only 3,211 followers were actively involved in the interaction, 1,695 followers were passive followers who did not like or did not comment on the interaction. Meanwhile, the results of the ayutingting92 follower cluster show that out of 1 million sample data followers, only 13,591 followers were actively involved in the interaction of likes and comments, 986,409 were passive followers.