基于用户脸部信息的推荐系统设计

Design of recommendation system based on user's face information

针对大数据时代的数据利用率不高情况,分析比较基于用户的协同过滤推荐算法和基于物品的协同过滤推荐算法;结合2种算法的优点,设计基于用户人脸信息的实时采集入库以及数据分析推荐系统,采用python语言编写程序功能模块。实验数据表明,在使用包含106 863张530名男女人脸图像的数据集训练和测试后,与传统有76. 6%的识别率的支持向量机(SVM)分类器模型、以及有94. 8%的识别率的融合局部二值模式(LBP)算法及SVM分类器算法的模型相比,在使用卷积神经网络(CNN)算法对该数据集构建模型则有98. 1%的识别率,相较前2种算法分别提升了21. 5%和3. 3%。因此,使用卷积神经网络算法训练数据集可以获得较高人脸检测及识别性别、年龄精度的模型。

In view of the low data-using efficiency situation in the era of big data, item-based collaborative filtering algorithm and user-based collaborative filtering algorithm are analyzed and compared, in which the face image could be gathered real-time;Python program language is used to conduct functional module program design in a data mining recommendation system which is the advantages of these two algorithms. The experiment results indicate that after using convolutional neural network (CNN) algorithm to training and testing by a dataset which contains 106 863 images of 530 different males and females,there is about 98.1% recognition rate. Compared with the traditional support vector machine (SVM) classifier model with a recognition rate of about 76.6%, and a fusion algorithm of LBP and SVM classifier algorithm with a recognition rate of about 94.8%, CNN algorithm model improves the face recognition rate by 21.5% and 3.3%. Therefore, the high accuracy of face detection and recognition model for two algorithms variables including gender and age, which could be accomplished by using convolution neural network algorithm to train data sets.