| In recent years,the e-sports industry has been developing rapidly.With the expansion of the e-sports market,more and more people have begun to engage in big data work of e-sports.In traditional competitive sports competitions,pre-match data analysis plays an extremely important role.In almost all the world’s top competitive sports competitions,there will be a dedicated data analysis team for data collection and data analysis to help the team win.The League of Legends professional game is currently the most watched e-sports game in the world.Its game mechanism is complicated,and the game is full of operational strategies and skills.Therefore,the League of Legends game is extremely enjoyable and it is difficult to predict the final winner.Based on the above reasons,constructing a League of Legends winning rate prediction model to help the coaching team perform pre-match analysis will have great research significance for improving the team’s winning probability.The main research work of this article includes the following parts:1)Word vector.In the task of winning percentage prediction,there are two relationships between collocation and restraint.According to the specific manifestation of the feature relationship,the Word2 Vec model is used to solve the problem of hero collocation relationship.On the basis of the word vector generated by the model,the restraint coefficient of each two heroes can be calculated through the restraint relationship algorithm,and then an expansion vector with restraint relationship is generated.2)Winning rate prediction model based on LSTM.Aiming at the problem of the front and back correlation between heroes in the input lineup,a winning rate prediction model based on LSTM is proposed.This model effectively improves the accuracy of the prediction model by extracting the features of the front and back positions in the lineup.3)Winning rate prediction model based on the improved BILSTM＿Att.Aiming at the different degree of influence of heroes on the results in actual games,the attention mechanism is used to assign the weight of heroes,and then the BILSTM＿Att winning rate prediction model is proposed.The model first uses the BILSTM layer to extract the location features of the input data,then uses the attention mechanism to assign feature weights,and finally uses softmax classifier to classify.4)Winning rate prediction model based on the improved CNN-BILSTM＿Att.According to the characteristics of the CNN model that can effectively extract the local features of the sample,in order to further optimize the prediction model,the CNN-BILSTM＿Att winning rate prediction model is proposed.The model first extracts the local features of the lineup data through the CNN layer,then uses the BILSTM layer to extract the front and back position features of the data,then uses the attention layer for weight distribution,and finally uses the softmax classifier to classify. |
