Lottery Sales Volume Forecast Model Based On Deep Learning

Since the issuance of Chinese lottery tickets in 1987,it has raised more than one trillion yuan in public welfare funds,which has effectively supported the improvement of people's welfare and the development of sports.Considering that lottery is a national public welfare undertaking,it is of great practical significance to study the factors affecting its sales volume,establish a lottery sales volume prediction model,and ensure the sustained and healthy development of lottery sales.In the traditional lottery sales forecasting research,due to the objective factors such as complex factors affecting lottery sales and the difficulty in obtaining lottery sales data,there are only quantitative investigations of one or more influencing factors.The traditional time series prediction and linear regression prediction are the main methods.The selected characteristic factors and research methods have certain limitations,and the prediction effect is not ideal.Therefore,this thesis proposes to use the deep learning regression model to predict lottery sales.The actual prediction effect of the model is simulated and verified.The main contents are as follows:(1)Firstly,this thesis analyzes the influencing factors that play a major role in the study of lottery sales volume based on provincial units and annual cycles.Considering the difficulty of data collection,seven macro and micro factors are selected as characteristic variables.The statistical correlation analysis indicates that there is a statistical correlation between the characteristic variables and lottery sales.(2)When using empirical formulas,it's difficult to achieve the optimal hidden layer nodes structure in the deep neural network regression model.To solve this problem,the maximum information coefficient is used to measure the correlation between the outputs of each hidden layer node and the outputs of the model,and the weaker correlation nodes are eliminated,which adaptively determines the number of hidden layer nodes.The simulation results indicate that the optimal model structure optimized by this algorithm is the same as the optimal model structure by multiple iteration tests based on empirical formulas,which verifies the feasibility of the algorithm and determines the optimal model structure.(3)The improved support vector regression proportional insensitive loss function is used to optimize the loss function of the deep neural network model.When using the gradient descent algorithm to update the weight,it is easy to fall into the local best advantage.Therefore,the improved group intelligence algorithm TLBO algorithm(Teach-Learn Based Optimization algorithm)is introduced to periodically update the weight matrix,which enhances the algorithm's ability to global search and jump out of the local optimal,and improves the prediction accuracy of the model.The convergence proof of the algorithm is obtained.The simulation results indicate that the prediction accuracy of the DNN-SVR regression model is significantly better than the linear regression model,the linear kernel support vector regression model,and the deep neural network regression model;the integrated models optimized by the TLBO algorithm are closer to the optimal model,and the prediction accuracy of the models is higher.The simulation results verify the effectiveness of the TLBO algorithm.In summary,this thesis takes lottery sales forecasting as the research direction,and uses deep neural network regression model to predict lottery sales,which has some reference value for the healthy and sustainable development of China's lottery industry.