Research on film data analysis and prediction based on ensemble learning and transfer learning

In recent years, with the development of machine learning research and the emergence of new results, more and more machine learning algorithms have been widely used in daily life. At present, most researches on machine learning algorithms are based on some data sets, and the information contained in the data is mined by using reasonable data analysis methods, which has certain reference value for the development of some industries. Film data is characterized by large amount of data, wide range of sources, and easy access. Based on the background of film data analysis and prediction, this paper studies the integrated learning method based on Stacking, the multi-source weighted TrAdaBoost movie score prediction method, and the movie score prediction model based on domain adaptive algorithm. Firstly, in the part of film data preprocessing and feature engineering, this paper takes IMDB film data set as the research object to carry out the research work of feature coding, constructing film comprehensive type influence index, director and actor rating, feature screening, and deleting strange film samples. Then, this paper studies the movie score prediction model based on the integrated learning method Stacking. The base learner of this model integrates multiple single machine learning models. The experimental comparison shows that the integrated learning model in the prediction of movie sample ratings can effectively reduce the prediction error and has certain advantages compared with a single machine learning model. Next, in order to solve the problem that a large number of movie score labels are missing in movie data samples in real scenes, which leads to the failure to train a machine learning model with good predictive score effect, this paper combines transfer learning theory to predict the score of the model. In this paper, an instance-based multi-source weighted TrAdaBoost transfer learning method is first used to study the Stacking integrated learning model. Based on the research content of the ensemble learning part, the integrated learning model is used as the base regressor to predict movie scores. At the same time, in order to verify the effectiveness of multi-source domain transfer learning, this paper further carries out a comparative study of multi-source domain transfer, single source domain transfer and no transfer. Based on the research of single-source domain migration, in order to further improve the movie score prediction performance of the model, this paper proposes a multi-source weighted TrAdaBoost model with multiple regressors, and verifies its effectiveness through comparative research. Finally, a feature-based domain adaptive transfer learning method is adopted to carry out the research. In the research, it is found that the feature dimensions of the target domain movie sample and the source domain movie sample are inconsistent, so the domain adaptive transfer learning model combined with dimensionality reduction algorithm is adopted in this paper. At the same time, in order to reduce the movie score prediction error of the model, this paper uses the integrated learning model as the regressor based on the integrated learning part of the research. Finally, through comparative experiments, the validity of the domain adaptive transfer learning model studied in this paper is verified, and it can better achieve the score prediction of the target domain movie samples.