A Comparative Analysis of Intelligent Classifiers for Mapping Customer Requirements to Product Configurations

Intelligent classifiers are now widely used to solve engineering problems. This study analyzes the performance of six different classifiers under various sparsity for mapping customer requirements to product configurations for the first time. The mapping procedure can be regarded as the classification task for selecting the product configuration which best satisfy the requirements from every customer. Data from two real word design cases are used to illustrate the comparison, the 10 times 10-fold cross-validation is utilized as evaluation model, the accuracy and macro F-measure are applied as performance metrics. The results of the two cases indicate that the Naive Bayes Classifier, Support Vector Machine and Relevance Vector Machine perform better than C4.5 Decision Tree, K-Nearest Neighbors and Multilayer Perceptron (a method of Artificial Neural Networks). This study is expected to serve as supportive research on the application of intelligent classifiers in product design issues regarding high dimensional sparse data.