Please enable JavaScript.
Coggle requires JavaScript to display documents.
Predicting Box-Office Success of Motion Pictures with Neural Networks -…
Predicting Box-Office Success of Motion Pictures with Neural Networks
Methodology
Classify the financial performance of a movie at the box office before its theatrical release using MLP neural networks..
Movie classification is based on its box office receipts in one of nine categories ranging from flop (lowest class) to blockbuster (highest class)
Model Architecture
The MLP NN, predicts the box office gross revenues (exclduing auxiliary revenues) before the movie's theatrical release.
The MLP includes apart from an input and output layers two hidden layers. In both hidden layers, sigmoid transfer functions were used
10-fold cross validation approach was also used to train the model on all but one fold (training set) then tested on the remaining fold (validation set)
Another feature of the proposed model is the use of the discretization concept, in which box office receipts are a limited number of intervals in a continuous spectrum, this is in order to find the appropriate class for a movie.
Results of Study
Results had shown that the MLP had on average 36.9% absolute accuracy, and 75.2% relative accuracy. To test the model's performance, other methods were used and results had shown that the proposed MLP had generated better classification accuracy than its competitors
Possible Enhancements / Future Works
The proposed NN model could be modified by adding other determinant variables to improve the NN accuracy.
The proposed model could be used in forecasting the success rates of a movie
Weaknesses
Dataset used in the proposed model does not take into account that more than one movie may share some of the same data e.g., actors, but the movies are categorized differently. To have more accurate results, the dataset needs to be modified and movies should not be dependent.
Inputs and Outputs of Model
Inputs
The study had included the following as inputs to the MLP, MPAA rating, competition, star value, genre, technical effects, sequel, number of screens.
Outputs
The output would be to classify the movie based on the inputs in one of the nine categories from 1 (flop) to 9 (blockbuster)
Sample
834 movies released between '98 and '02
Comparison to Other Models
Multinomial Logistic Regression (logit)
Average results:
Relative accuracy: 69.60%
Absolute accuracy: 30.17%
Discriminant Analysis
Average results:
Relative accuracy: 67.69%
Absolute accuracy: 29.25%
Classification and Regression Tress (CART)
Average results:
Relative accuracy: 71.07%
Absolute accuracy: 31.18%
MLP (Proposed Model)
Average results:
Relative accuracy: 75.20%
Absolute accuracy: 36.90%