Please enable JavaScript.
Coggle requires JavaScript to display documents.
Predicting Movie Box-Office Revenues Using Deep Neural Networks - Coggle…
Predicting Movie Box-Office Revenues Using Deep Neural Networks
Methodology
To develop a multimodal DNN, while also incorporating input about movie poster features learned in a data-driven fashion, for movie box office revenues prediction before the theatrical release of the movie. The author argues that it is expected that the extracted features from movie posters may have an impact on the performance of the DNN.
Model Architecture
1- A CNN would be trained first on learning and extracting features from movie posters.
2-Then, CNN is used as a module in the multimodal DNN. The movie poster content extracted by the CNN is used with other movie-related data as input.
Possible Enhancements / Future Works
Using this model, audio and video data from movies can be used as inputs by building additional multimodal DNNs.
Weaknesses
Accuracy results show a good sign of the use of multimodal DNN as a prediction system, however, accuracy can be increased by adding more layers to the CNN to help extract more features from the movie posters but up to a certain extent to avoid over fitting.
Inputs and Outputs of Model (DNN)
Inputs
Features from the movie poster (done by the CNN, and this step is done first before proceeding to the prediction "DNN")
Genre
Duration or running time of movie
Production budget
Movie participants
Movie critique
Consumer rating
Output
Movie box-office revenues prediction
Sample
3807 movies were selected
Comparison to Other Models
Excluding Movie Posters
RF
Average results:
Relative accuracy: 82.23%
Absolute accuracy: 46.38%
MLBP
Average results:
Relative accuracy: 84.30%
Absolute accuracy: 48.89%
SVM
Average results:
Relative accuracy: 62.67%
Absolute accuracy : 33.15%
Proposed DNN
Average results:
Relative accuracy: 86.44%
Absolute accuracy: 51.45%
Including Movie Posters
RF
Average results:
Relative accuracy: 82.86%
Absolute accuracy: 47.78%
MLBP
Average results:
Relative accuracy: 85.93%
Absolute accuracy: 50.06%
Proposed DNN
Average results:
Relative accuracy: 88.60%
Absolute accuracy: 52.20%
SVM
Average results:
Relative accuracy: 64.84%
Absolute accuracy: 34.37%
Results of Study
Two experiments were conducted to evaluate the impact of movie posters usage as input, first excluding the posters, then including them.
The multimodal DNN had shown better results than its competitive methods (SVM, RF, MLBP) in both experiments, which shows first the proposed NN has a better prediction performance and the usage of movie posters has an impact on the prediction outcome as there is an increase in the performance.