Box-office forecasting based on sentiments of movie reviews and Independent subspace method
Citations
109 citations
71 citations
Cites methods from "Box-office forecasting based on sen..."
...[12] employed multiple linear regressions, classification and regression trees, artificial neural networks, and support vector regression models to forecast the number of films audiences....
[...]
68 citations
52 citations
30 citations
References
40,147 citations
"Box-office forecasting based on sen..." refers methods in this paper
...The performance variation from worst to best among the forecasting models when the ISM is applied is about -5% to 15% in the ANN, and about -8% to 13% in the SVR, respectively, which is smaller than the approximately -43% to 12% in the CART....
[...]
...Model W23 (j = 2, k = 3) Category Variable MLR CART ANN SVR Total External Factor Ratio Scrk 86.53∗∗∗ 102.02∗∗∗ 80.89∗∗∗ 94.61∗∗∗ 4 Seasonalityk 5.47∗∗∗ 4.69∗∗∗ 2.86∗ 6.23∗∗∗ 4 Audience Factor Avg ratingj 4.49*** 2.15∗ 7.31∗∗ 6.02∗∗∗ 4 N ratingj 0.22∗ 1.38∗∗∗ 2 Sum PosRevj 0.24*** 7.88∗∗∗ 1.24∗∗ 3 Sum NeuRevj 0.24∗ 1 Sum NegRevj 1.17∗∗∗ 1.39*** 2 Ratio PosRevj 0.24*** 0.07 2 Ratio NeuRevj 0 Ratio NegRevj 0.12∗∗∗ 0.96 0.27∗ 3 Motion Picture Factor KOR 0.09 1.35∗∗∗ 2 US 0 0.15 0.75*** 3 MPAA 0.56∗∗∗ 0.96∗ 1.42∗∗∗ 3 Dir Pop 0.09 0.54 0.64∗∗∗ 3 Sales Pow 0.01 1 Dis Pow 1.34*** 1 Act Pop 2.84∗∗∗ 1.00∗∗ 2.87 4.48∗∗∗ 4 N audj 388.36∗∗∗ 433.03∗∗∗ 373.41∗∗∗ 397.70∗∗∗ 4 Total number of important factors 13 8 8 17 week, which is considered as the most influential variable for the box-office forecasting, is included or not....
[...]
...Algorithm Parameters Settings MLR - - CART Pruning Off Min. number of observations per tree leaf 1 Min. number of observations per tree parent 10 ANN Number of hidden nodes 1, 3, 5, 7, 9, 11, 13, 15 Max. number of epochs 300 Target training error 1.0× 10−7 SVR Kernel type Gaussian (RBF kernel) Gamma 2 ∧ {−10,−9, · · · ,−1, 0} nu 0.1, 0.3, 0.5, 0.7, 0.9 Cost 2 ∧ {−1, 0, · · · , 7, 8} instances were randomly sampled....
[...]
...Model W01 (j = 0, k = 1) Category Variable MLR CART ANN SVR Total External Factor Ratio Scrk 197.07*** 347.81*** 273.68*** 247.78*** 4 Seasonalityk 27.84*** 20.67*** 34.05*** 39.81*** 4 Audience Factor Avg ratingj 6.98*** 1.87* 7.98*** 3.24*** 4 N ratingj 1.47*** 1.23 4.11*** 5.83*** 4 Sum PosRevj 0 Sum NeuRevj 0.11 0.97 1.96 0.45 4 Sum NegRevj 0.2 1 Ratio PosRevj 0.72 1.32 0.17 3 Ratio NeuRevj 3.71** 1 Ratio NegRevj 7.90*** 0.59*** 2 Motion Picture Factor KOR 2.05*** 3.32** 4.48*** 3 US 3.03*** 1.46* 7.06*** 4.19*** 4 MPAA 1.37*** 1.03 9.22*** 3 Dir Pop 3.70*** 4.74*** 3.87*** 6.17*** 4 Sales Pow 1.77* 1 Dis Pow 0.88*** 0.65* 2.37* 0.35 4 Act Pop 17.08*** 17.33*** 21.59*** 15.43*** 4 N audj − − − − − Total number of important factors 11 11 15 13 prior to the release....
[...]
...The best algorithm from the ANN and SVR varies depending on the forecasting model and evaluation criteria....
[...]
10,696 citations
7,452 citations
"Box-office forecasting based on sen..." refers methods in this paper
...A sentiment corpus was built by simply looking up a small 135 number of words in a dictionary in the public domain [41] because building up a large number of words for each domain requires significant time and cost [17, 42, 54]....
[...]
5,664 citations
4,526 citations
"Box-office forecasting based on sen..." refers background in this paper
...A particular expression can be viewed as positive or negative based on the domain [49, 52]....
[...]