Implementation Of Naive Bayes Algorithm With Particle Swarm Optimization In Classification Of Dress Recommendation

Fashion is considered as a cycle of reflection of unique social, cultural and environmental characteristics in time besides playing an important role in completing one's self-image. Current developments in technology and information are foremost for the growth in the amount of data held been recorded and stored in large databases (data mountains) to turn into a piece of knowledge by classifying using machine learning. One of the classification methods that can be used is Naïve Bayes. The purpose of this study is Dress Classification Recommendation using the Particle Swarm Optimization-based Naïve Bayes algorithm by increasing the existing accuracy value. This study uses a public Dresses Attribute Sales dataset downloaded at the UCI Machine Learning repository with 13 Attributes and has two labels, 1 for recommendation and 0 for no recommendation. The amount of data is 500 data. The evaluation used is to use the Confusion Matrix, to determine the value of accuracy, precision, and recall. The results of the experiments carried out by using the optimization feature particle swarm optimization with population size 20 parameters and maximum of number generation 35, then using Cross-Validation with Naïve Bayes algorithm. The value of Cross-Validation used is 3, 5, 7 and 10, that the highest value of accuracy is to use the value of Cross-Validation (K Fold) K 10, with an accuracy value of 65.40% with precision 67.28% and recall 80.34%.