No abstract

Abstract. In order to detect and identify fruit category more efficiently, this study presents a novel method based on Haar wavelet entropy, multilayer perceptron, and standard genetic algorithm. The Haar wavelet entropy extracted features from a given fruit image. The multilayer perceptron received the features and acted as a classifier. Finally, genetic algorithm was used to train the classifier. The experiment was performed over a 10x12-fold cross validation. The overall accuracy was 81.11±4.23%., better than the result of back propagation gradient descent algorithm of 74.17± 4.98%, and the result of simulated annealing of 78.10± 3.57%.

Abstract-In order to predict physics achievement in middle school, this paper proposed a new method based on big five model. First, we collected 300 samples, in which 150 passed and the other 150 failed the final physics examination. Then, we submitted the five demographic features and five big-five personality trait features to the artificial neural network (ANN). Third, we used back propagation algorithm to train the ANN. The cross validation results show that our method yielded a sensitivity of 83.00± 2.09%, a specificity of 82.73± 4.12%, and an accuracy of 82.87± 2.75%.