Comparison of Optimization Using Latent Semantic Analysis and Latent Dirichlet Allocation on Aspect-Based Sentiment Analysis of Online Food Delivery (OFD)

Sentiment analysis offers the food and beverage industry better business outcomes. It is necessary to understand customer opinions based on the services obtained. However, sentiment analysis is still not enough to provide knowledge of what specific things stand out and are lacking. The need for aspects sold by restaurants, namely the 4 P (Product, Price, Place, Promotion) Marketing Mix. Aspect-Based Sentiment Analysis will be performed on customer reviews of restaurant with OFD (Online Food Delivery). The work process will be based on the CRISP-DM framework with the support of an understanding of the value in the form of 4Ps sold by OFD restaurants. Scraping is done on the website and processing the scraping results by doing text pre-processing. Furthermore, model building is carried out by comparing the Support Vector Machine algorithm added with Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA). Data was taken from April 2024 with a total of 2820 reviews. Based on the model that has been created, the LDA model is better than the LSA model with an accuracy benchmark, namely a difference of 3 percent.