forecast model for box-office revenue of bollywood feature films

Post on 05-Apr-2017

769 Views

Category:

Technology

4 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Forecast Model for Box-Office Revenue of Bollywood Feature Films using Machine Learning

B. E. Computer EngineeringNetaji Subhas Institute Of Technology,

New Delhi

March 13, 2015

Presented by:

Prerit Kohli

PGP at Indian Institute of Management, Indore

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 2/33

PROBLEM STATEMENT

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 3/33

Problem Statement

The aim is to forecast the Box-office revenue for a Bollywood feature film using Machine Learning, by adding the computed influence of each parameter of a movie that is believed to affect its revenue.

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 4/33

MOTIVATION

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 5/33

Motivation

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Bollywood is the world’s largest filmmaking entity, with over 1,000 films produced annually.

Bollywood generated revenue of around Rs. 15,000 crores in 2011 and this figure has been growing by 10 percent a year.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 5/33

Motivation

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Bollywood is the world’s largest filmmaking entity, with over 1,000 films produced annually.

Bollywood generated revenue of around Rs. 15,000 crores in 2011 and this figure has been growing by 10 percent a year.

It has a range of attributes such as the Music-album industry and the “masala” film genre, distinct from film industries in other countries.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 6/33

Motivation

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Unlike Hollywood, much research has not been done on forecasting for Bollywood feature films.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 6/33

Motivation

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Unlike Hollywood, much research has not been done on forecasting for Bollywood feature films.

Forecast model used to assist film studios as even a single movie can be the difference between crores of rupees of profit or loss in a given year[1].

[1] Jeffrey S. Simonoff, Ilana R. Sparrow (2000), Predicting movie grosses: Winners and losers, blockbusters and sleepers. Stern School of Business, New York University.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 6/33

Motivation

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Unlike Hollywood, much research has not been done on forecasting for Bollywood feature films.

Forecast model used to assist film studios as even a single movie can be the difference between crores of rupees of profit or loss in a given year[1].

Forecast model also used to assist cinema hall/multiplex owners in planning out movie schedules for forthcoming box-office weekends.

[1] Jeffrey S. Simonoff, Ilana R. Sparrow (2000), Predicting movie grosses: Winners and losers, blockbusters and sleepers. Stern School of Business, New York University.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 7/33

METHODOLOGY

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 8/33

Methodology

Pre-production

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Fig 1 Lifecycle of a Feature Film

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 8/33

Methodology

Pre-productionFilm shoot & dubbing

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Fig 1 Lifecycle of a Feature Film

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 8/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Fig 1 Lifecycle of a Feature Film

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 8/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production Release

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Fig 1 Lifecycle of a Feature Film

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 8/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production Release Post-release

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Fig 1 Lifecycle of a Feature Film

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 9/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production Release Post-release

Post-production Method

Ek VillainJune 27, 2014

1

1

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 9/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production Release Post-release

Post-production Method Next-change Method

Ek VillainJune 27, 2014

HolidayJune 6, 2014

2

1 2

1

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 9/33

Methodology

Pre-productionFilm shoot & dubbing

Post-production Release Post-release

Post-production Method Post-release MethodNext-change Method

Ek VillainJune 27, 2014

CityLightsMay 30, 2014

HolidayJune 6, 2014

2 3

1 2 3

1

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 10/33

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

IMPLEMENTATION

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 11/33

Regression Analysis

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Regression is the Machine Learning technique used in our forecast model.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 11/33

Regression Analysis

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Regression is the Machine Learning technique used in our forecast model.

Datasets of parameters of already-released Movies are built and fed to the machine.

Actual revenues of the movies are also fed.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 11/33

Regression Analysis

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Regression is the Machine Learning technique used in our forecast model.

Datasets of parameters of already-released Movies are built and fed to the machine.

Actual revenues of the movies are also fed.

The machine learns from these datasets, the influence of each parameter on the movie revenue.

This analysis is used to forecast revenues for upcoming movies.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 12/33

Linear Regression

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

The Linear Regression formula is as follows:

R = β1(P1) + β2(P2) + β3(P3) +... + βn(Pn)

where R is the forecasted revenue for the film, Pn is the value of nth parameter for the film, and βn is the corresponding coefficient of the nth parameter[2].

[2] Jae-Mook Lee, Tae-Hyung Pyo, Forecast Model for Box-office Revenue of Motion Pictures, Dec 2009.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 13/33

Post-production Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied when the film is completed and sent to the studio.

Used by Production houses (eg. Yash Raj Films) for deciding the marketing budget of an upcoming movie.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 13/33

Post-production Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied when the film is completed and sent to the studio.

Used by Production houses (eg. Yash Raj Films) for deciding the marketing budget of an upcoming movie.

Following parameters are considered:Top Actor/Actress Trending Actor/Actress

Top Director Promising Director

Sequel /Trilogy Top Production House

Movie Genre Movie Budget

Adaptation/Remake buzz Success record of Cast/Crew

Table 1 Parameters for Post-production Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 14/33

Post-production Method: Average Error

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Fig 2 Average-error plot for Post-production Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 15/33

Next-change Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied when the movie prints are sent to the theaters a few days before the release date.

Used by movie exhibitors (cinema halls) for finalizing on the number of shows to be devoted to an upcoming movie.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 15/33

Next-change Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied when the movie prints are sent to the theaters a few days before the release date.

Used by movie exhibitors (cinema halls) for finalizing on the number of shows to be devoted to an upcoming movie.

It adds the following parameters of its own, along with those of previous method:Music-album popularity Movie Screens across India

Out-of-budget promotion Critics Reviews from Paid-previews

Censor Board Rating (U/UA/A) Competition from movies sharing same release-date

Table 2 More parameters added for Next-change Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 16/33

Next-change Method: Average Error

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Fig 3 Average-error plot for Next-change Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 17/33

Post-release Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied at the end of the first weekend of the release-date.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 17/33

Post-release Method

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Applied at the end of the first weekend of the release-date.

This method adds the following parameters of its own, along with those of the Post-production and Next-change Methods:

Critics’ Reviews Audience Response

Unexpected promotion post-release Promotion by Govt. (E-tax exemption)

Viral word-of-mouth

Table 3 More parameters added for Post-release Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 18/33

Post-release Method: Average Error

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Fig 4 Average-error plot for Post-release Method

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 19/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Trending Actor (m): Arjun Kapoor (0.32)

Trending Actor (f): Alia Bhatt (0.23)

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 19/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Trending Actor (m): Arjun Kapoor (0.32)

Trending Actor (f): Alia Bhatt (0.23)

Estb. Production House: Dharma Productions (0.22)

Budget: Rs. 36 Crores (0.28)

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 19/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Trending Actor (m): Arjun Kapoor (0.32)

Trending Actor (f): Alia Bhatt (0.23)

Estb. Production House: Dharma Productions (0.22)

Budget: Rs. 36 Crores (0.28)

Adaptation: Chetan Bhagat’s “2 States” (0.25)

Music album popularity: Very good response (0.24)

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 19/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Trending Actor (m): Arjun Kapoor (0.32)

Trending Actor (f): Alia Bhatt (0.23)

Estb. Production House: Dharma Productions (0.22)

Budget: Rs. 36 Crores (0.28)

Adaptation: Chetan Bhagat’s “2 States” (0.25)

Music album popularity: Very good response (0.24)

Genre(s): Drama (-0.1) + Romance (0.21)

Censor Board rating: U/A (0.32)

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 20/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Predicted Revenue: R

Log10 R = 0.32 + 0.23 + 0.22 + 0.28 + 0.25 + 0.24 + (-0.1) + 0.21 + 0.32= 1.97

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 20/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Predicted Revenue: R

Log10 R = 0.32 + 0.23 + 0.22 + 0.28 + 0.25 + 0.24 + (-0.1) + 0.21 + 0.32= 1.97

Antilog(1.97) = 93.325Predicted Gross: R = 93.33 Crores

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 20/33

Revenue forecast for 2 States [2014]

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

2 States [2014] References

Regression Analysis

Post-production Method

Next-change Method

Post-release Method

Predicted Revenue: R

Log10 R = 0.32 + 0.23 + 0.22 + 0.28 + 0.25 + 0.24 + (-0.1) + 0.21 + 0.32= 1.97

Antilog(1.97) = 93.325Predicted Gross: R = 93.33 Crores

Actual Gross: 104.04 Crores

Percentage Error = |93.33 – 104.04|/104.04 = 10.29%

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 21/33

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

RESULTS

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 22/33

Results

Problem Statement Results

Motivation Challenges Faced

Methodology Learning Experience

Implementation Future Work

References

The following are the Average errors in the 3 methods adopted:

Post-production Method: 36.05%

Next-change Method: 19.52%

Post-release Method: 11.74%

This depicts the correlation between the numbers of revenue-affecting parameters and the accuracy of the revenue forecast.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 23/33

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

CHALLENGES FACED

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 24/33

Challenges Faced

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

> Forecasting revenues for surprise blockbusters such as Queen [2014]

> Incorporating multiple genres for movies.

> Demarcation for Database Lists. Doubts such as whether to include Sanjay Dutt in the Top Actors list.

> Computation of loss of revenue for movies sharing the same release-date.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 25/33

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

LEARNING EXPERIENCE

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 26/33

Learning Experience

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

> Current trends in the multi-billion Bollywood industry.

> Tapping Machine learning techniques for forecasting revenues of films we see every week.

> Strategies adopted by Film Studios and Film Exhibitors for maximum revenue generation.

> Statistical verification and graph plotting.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 27/33

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

FUTURE WORK

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 28/33

Future Work

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

> Sub-categorizing the Database for inclusion of prominent actors such as John Abraham.

> Incorporating the foreign Box-office of a film.

> Exploring more factors that determine movie revenues.

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 29/33

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

REFERENCES

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 30/33

References[1] Jeffrey S. Simonoff, Ilana R. Sparrow (2000), Predicting movie grosses: Winners and losers, blockbusters and sleepers. Stern School of Business, New York University.

[2] Jae-Mook Lee, Tae-Hyung Pyo, Forecast Model for Box-office Revenue of Motion Pictures, Dec 2009.

[3] Chrysanthos Dellarocas, Xiaoquan (Michael) Zhang, Neveen F. Awad. (2007, Aug.). Exploring the value of online product reviews in forecasting sales: The case of motion pictures. Journal of Interactive Marketing. [Online]. Available: http://blog.mikezhang.com/files/movieratings.pdf.

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 31/33

References[4] Nikhil Apte, Mats Forssell, Anahita Sidhwa, Predicting Movie Revenue, Dec 2011.

[5] Mahesh Joshi, Dipanjan Das, Kevin Gimpel, Noah A. Smith, Movie Reviews and Revenues: An Experiment in Text Regression. Language Technologies Institute, Carnegie Mellon University.

[6] Alec Kennedy, “Predicting Box Office Success: Do Critical Reviews Really Matter?”, The University of California, Berkeley.

[7] Márton Mestyán, Taha Yasseri, János Kertész (2013, Aug.). Early Prediction of Movie Box Office Success Based on Wikipedia Activity Big Data. Institute of Physics, Budapest University of Technology and Economics.

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 32/33

References

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

[8] “Movie Database", Available at https://www.imdb.com/

[9] “Box Office Revenue Database", Available at https://www.koimoi.com/

[10] “Music Popularity Index Database", Available at https://www.top10bollywood.com/

[11] “Film Critics Database", Available at https://www.hindustantimes.com/entertainment/https://www.ibnlive.in.com/movies/reviews/ https://www.bollywood.bhaskar.com/reviews/ https://zoomtv.indiatimes.com/ https://www.bollywoodhungama.com/reviews/

Prerit Kohli, Rajat Taneja, Saumya Bansal Movie Revenue Forecasting 33/33

Thank You

Overview Results

Objective Challenges Faced

Motivation Learning Experience

Methodology Future Work

Experiment References

top related