Abstract: In order to transform the usage data into appropriate information that can improve the products through design modification, a method based on convolution neural network is proposed for user experience evaluation modeling, which can make full use of the usage data to establish the mapping relationship between the user information and the product engineering requirements. Firstly, the time-series usage data was converted into a series of data units by sliding window technique, and a convolution neural network architecture suitable for user experience evaluation model was established. Then, the optimal hyper parameters was selected and the over fitting problem of the model was improved by K-fold cross validation analysis. Finally, the validity of the proposed method was demonstrated by a case study of smart phone user experience evaluation modeling. The results indicated that the proposed method can automatically extract effective features from raw usage data, which can used for user experience evaluation prediction. Thus, the proposed method can decrease the dependence of the users and designers when modeling, which can help designers to assess the product performance in real time and accurately and provide support information for design decisions through usage data.

Key words: product design; user experience; usage data; convolutional neural network; feature extracting