Research And Design Of Film And Television Recommendation System Based On Fusion Deep Trees And Deep Learning

In recent years,with the rise of global trade protectionism and the intensification of conflicts,the traditional film and television industry has ushered in a cold winter.Under this background,online film and television platforms have developed rapidly,and massive film and television data have flooded into the platform,making it increasingly difficult for users to obtain the cultural products they need in a short time.In order to solve the difficulty of accurate recommendation on film and television platforms,this paper adopts an improved recommendation model based on deep learning.By taking advantage of the good adaptability of the deep learning model,the interactive layer with changing vector direction is designed to transform the traditional model.At the same time,the deep tree-based mechanism is introduced to conduct preliminary screening of data from massive candidate sets before embedding layer,and the initial data is cleaned to reduce noise.Subsequently,the model uses the idea of convolution to reduce the feature dimension,change the dilemma of high cost and high complexity caused by excessive reliance on artificial feature engineering,and improve the overall expressive ability of the model.The main research content of this paper is divided into the following parts.(1)Study the structure,advantages and disadvantages of Wide&Deep models based on deep learning,improve the derived model DCN,which has serious information loss after the combination of feature dimension reduction.Combine the idea of convolution kernel in computer vision and other fields,replace the Cross Network module of the model,and enhance the model’s feature expression ability.Finally,several models are compared on a data set to verify the effect of the improved model.(2)In order to further enhance the initial screening ability of the model to cope with massive data candidate sets,the deep tree mechanism was introduced to study a recommendation system model based on the fusion of deep tree and deep learning.Combined with the improved Beam Search algorithm,the depth tree retrieval was generated to reduce the proportion of noise data before the initial feature input of the model.Finally,comparative experiments were conducted on offline and online data sets to analyze indicators such as accuracy rate,recall rate,AUC and Logloss,and to compare the influence of several hyperparameters on the model to verify the model effect.(3)Design a film and television recommendation system based on the model proposed earlier.Introduce the main process of system design,including requirement analysis,system framework design,and design the overall structure of each module and key databases based on the requirements.Finally,demonstrate the implementation effect of the system,including the front-end effect of the system and the recommendation function for film and television works.