British Journal of Computer, Networking and Information Technology

ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

MOVIE SUCCESS PREDICTION USING DATA MINING

Olubukola D. Adekola1, Stephen O. Maitanmi2, Funmilayo A. Kasali3,
Ayokunle Omotunde2, Oyebola Akande2,
Oduroye Ayorinde4, Wumi Ajayi2 and Yaw Mensah2
1
Department of Software Engineering, Babcock University, Ilisan-Remo, Ogun State,
Nigeria. E-mail: adekolao@babcock.edu.ng
2
School of Computing and Engineering Sciences, Babcock University, Ilisan Remo, Ogun
State, Nigeria
3
Department of Computer Science and Mathematics, Mountain Top University, Ogun State,
Nigeria
4
Computer Science Department, Caleb University, Imota, Lagos, Nigeria

Cite this article: ABSTRACT: The movie industry is arguably one of the biggest
Olubukola D.A., Stephen entertainment sectors. Nollywood, the Nigerian movie industry
O.M., Funmilayo A.K., produces tons of movies for public consumption, but only a few
Ayokunle O., Oyebola A., make it to box-office or end up becoming blockbusters. The
Oduroye A., Wumi A., Yaw
M. (2021), Movie Success introduction of movie success prediction can play an important
Prediction Using Data Mining. role in the industry not only to predict movie success but to help
British Journal of Computer, directors and producers make better decisions for the purpose of
Networking and Information
Technology 4(2), 22-30. DOI: profit. This study proposes a movie prediction model that
10.52589/BJCNIT- applies data mining techniques and machine learning
CQOCIREC. algorithms to predict the success or failure of an upcoming
movie (based on predefined parameters). The parameters
Manuscript History needed for predicting the success or failure of a movie include
Received: 30 Aug 2021 dataset needed for the process of data mining such as the
historical data of actors, actresses, writers, directors, marketing
Accepted: 17 Sept 2021
and production budget, audience, location, release date, and
Published: 22 Sept 2021
competing movies on same release date. This model also helps
movie consumers to determine a blockbuster, hit, success rating
Copyright © 2020 The Author(s). and quality of upcoming movies before deciding on a movie
This is an Open Access article
distributed under the terms of ticket. The data mining techniques was applied to Internet
Creative Commons Attribution- Movie Database MetaData which was initially passed through
NonCommercial-NoDerivatives
4.0 International (CC BY-NC-ND cleaning and integration process.
4.0), which permits anyone to
share, use, reproduce and KEYWORDS: Data Mining, Prediction, Machine Learning,
redistribute in any medium,
provided the original author and Movies, Meta Data
source are credited.

22 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

INTRODUCTION
In entertainment sector, movie industry is one of the biggest and most sought after. It is no
surprise that technology-based startups like Netflix, Irokotv and Hulu are on the rise with the
aim of expanding the reach of movies. However, only a small percentile achieves success and
high ratings while a lot suffer low viewership, some never earn enough profit to cover their
budget, while others never make it to box-office or attain considerable success. A movie
budget covers various areas like marketing campaign, operational cost, videography, payment
of team members and crew (Wales, 2017). If the movie ticket sales are lower than the cost of
making the movie then the movie is a failure. Theatre or media that cannot make profit is
bound to eventually fizzle out of existence (Ogunleye, 2004). A movie revenue is dependent
on a number of factors namely cast of the movie (popular actors have propensity of
increasing viewership), budget for the movie, budget for marketing and media campaign, film
critics review, ratings and release date. While there might be no exact formula to calculate or
provide analysis for predicting in value terms what a movie will generate even with all the
parameters listed, data mining could be used to analyze dataset from previous movies, which
can include revenue generated from pervious movies, release date and crews to predict the
success of an upcoming movie. The use of data mining approach is very potent as it helps in
identifying hidden patterns and relationship among various variables (Shu, 2020). The
introduction of data mining techniques in predicting movie success would mean developing a
model that could be used for future prediction. The technology that would help existing
entertainment businesses thrive better by assisting better decision making before and during
movie production is worth careful study. Whenever the proposed model is able to predict
movie success rate with at least 97-98% accuracy, it means higher profits can be generated.
The proposed study aims to develop a model based on data mining techniques that would be
applied in predicting the success of upcoming movies. The use of historical data of movies to
predict the success of upcoming movies such as director, crew and budget reduces certain
level of uncertainty (Nithin, Pranav, Sarath Badu & Lijiya, 2014).
Problem Statement
Outside of the crew’s popularity and movie popularity, the success of a movie is highly
dependent on digital forces like online campaign, marketing, reviews, box office ratings and
opening day ticket sales. However, before a movie is released, the success of the movie has
to rely on media hype, previews and prelease marketing campaign, but these do not
determine or translate to a movie’s success when released. The problem here is most
producers, directors and stakeholders end up expending millions on movie budget without
knowing if the movie would be a success or a failure. This study proposes a model that
analyzes the revenues generated from previous movies, the reviews, ticket sales, crew’s
popularity and marketing budget to predict the success of a movie. Such a prediction could
be very useful for motion studios to make better intelligent decisions like improving content
and creativity, artist compensations, advertising and marketing of the movie accordingly.
This helps investors and stakeholders to predict an accurate return-on-investment (ROI) and
other associated profit or loss.

23 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

LITERATURE REVIEW
The term Artificial intelligence (AI) was originally coined by John McCarthy in 1956 at a
conference on the subject (McCarthy, 2000). AI is a wide aspect of computer science that
deals with conception of intelligent machines that function like humans in workings and
reactions (Grewal, 2014). Data mining is an aspect of AI that involves data processing using
sophisticated data search capabilities and statistical algorithms to discover patterns and
correlations in large preexisting databases; a way to discover new meaning in data (Han,
Kamber & Pei, 2011). It helps in discovering patterns, usually difficult to find, in order to
decide upon the future trends in businesses. Data mined is traditionally viewed as a model of
the semantic make-up of the dataset, where the model might be utilized on new information
for forecast or classification. The main goal of data mining is to detect patterns that are
previously unknown and further use them to make informed decisions for development of
systems or models that improve businesses, decision making process and analysis (Ramageri,
2010). Movie data mining has employed methods as data processing, relationship mining,
prediction and clustering. Clustering is an unsupervised method that breaks data into definite
classes where the data in the same category are alike in properties but are different from other
data in other clusters. Clustering plays a significant role especially in data mining
applications, such as marketing, text mining, medical diagnostics, and so on (Mesakar, &
Chaudhari, 2013). Clustering in movie data mining searches for resemblances and variations
between movie and organizes them under similar categories for the purpose of movie success
prediction. Discoveries with models may be applied well in detecting relationships among
movies and movie success features or related variables. In movie data mining, predictions are
used to study features of a model that significantly predicts crucial information on the
essential construct. Distillation of data for human judgment focuses on making information
comprehensible. Preparation process for building models uses Data distillation for
classification while identification displays data in such a way that makes it simple to
recognize with a well-known pattern (Baker, 2010). Relationship Mining (RM) operates with
a focus of unearthing a huge number of variable connections among the variables in a dataset.
In movie data mining, there are four frequently used types of RM; association rule mining,
sequential pattern mining, correlation mining, and causal data mining. Relationship Mining is
applied within the aspect of finding the link that depicts connections of the data of related
movies based on certain features. This helps in enhancing the prediction model.
The three common learning types in Machine Learning (ML) approaches include supervised,
unsupervised and reinforced learning. The learning process in supervised learning model are
training and testing. The process of training involves taking samples in training data as input
in which features are learned by the learning algorithm or learner and building the learning
model (Dhage & Raina, 2016). The execution engine is used in the testing process by the
learning model in order to predict the test or production data. The unsupervised learning type
is of high importance within the process of multimedia content because partitioning of data
when class labels are not involved is mostly a necessity (Greene, Cunningham & Mayer,
2008). Unsupervised ML techniques ease analysis of fresh datasets in a way that helps in
creating analytic insights from unlabeled data (Usama et. al., 2017). There are various
applications from unsupervised Machine Learning like feature learning, data clustering,
dimensionality reduction, anomaly detection etc. In reinforced ML, there is an exposition of
the machine to an environment where they get trained through trial and error. Reinforced
learning helps to gain knowledge from the past experiences and uses it to get the most

24 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

appropriate knowledge in making appropriate choices; that is, based on the feedback received
(Kumar, Ramakrishnan & Li, 2018).

MATERIALS AND METHODS

The methods capture different stages of model development consisting data collection, data
preprocessing, generating training and testing dataset, model generation, prediction and
outcomes. This helps to eliminate irrelevant data in order to enhance accurate prediction.
Following this due process aids full cleaning of the dataset and discard of irrelevant data from
the IMDB dataset as well as through detailed study of the dataset; only the attributes that
could affect the prediction of success of a movie were used. The textual data was transformed
into numeric data and converted into CSV (comma separated version) format. Most suited
algorithm which yielded the highest accuracy and least error was employed. The model was
developed using Python, IMDB dataset and python libraries. The model used historical data
in order to successfully predict the possibility of success or failure of an upcoming movie.
Data mining techniques were used for analyzing patterns in previous records of movie data
and sales figures upon which classification was based.
The following (Figure1) represents the flow of activities in the system design:

START

Data Splitting

Data Pre-processing

Training
Train Model
Data Set

Test Model Evaluation

Data Set

Prediction
New Movie Movie Prediction Model
Data
Data
Movie Success Outcome

END

Figure 1: Diagram Illustrating the Flow of Activities in the System Design

25 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

Data Collection

Secondary data were pulled from online open source (IMDB) and used for training the
predictive model. The structure of the available data is illustrated below:
Table 1: Structure of data fetched for the predictive model

S/N VARIABLE NAME VARIABLE FORMAT VARIABLE TYPE

1 IMBD id Characters Continuous
2 Title Characters Continuous
3 Rating Float
4 Directors Characters Continuous
5 Writers Characters Continuous
6 Actors Characters Continuous
7 Budget Int Continuous
8 Gross Int Continuous
9 Financial Score Float Continuous
10 Genre Characters Categorical

Model Development
Data collection and cleaning are critical input for the successful development of any machine
learning model. As data were extracted from online repository, irrelevant variables and
features are eliminated. Cleaning of the dataset is aimed at removing inconsistences and other
anomalies. Data analysis is done after the pre-processing stage. The data is partitioned into
two sets. The main part of the dataset consists of 80% of the data used in the building of the
model using appropriate machine learning algorithm. The remaining part of the dataset is
used to train the model. Classification algorithms were deployed to make the built model
predict the success of upcoming movies.
Selection of Classification Algorithm
Engaging the use of multiple models in Machine Learning to determine the one that works
best for any set task (Cai, 2014) is the norm as no one algorithm can work best in any given
situation according to the No Free Lunch theorem (NFLT) (Wolpert & Macready, 1997).
Application of this rule is usually seen under Supervised Learning as validation and cross-
validation are common techniques deployed in assessing the predictive accuracies of multiple
models to determine which is best with regards to the task at hand. Training models that work
well with multiple algorithms is also good practice. The classification algorithms chosen to
train the models include Decision Trees, Gradient Boosting, Naïve Bayes and Multi-layer
Perception (MLP).
Model Validation
This measures the accuracy of the data mining system against real data. This is applied to
verify if the developed system built on training set actually has the capacity to handle general
data (Alpaydin, 2010). Applying model validation ensures that the system avoids over-fitting.
Some of the cross-validation techniques that can be used include hold-out method, leave-out
validation technique and k-fold cross validation. K-fold validation would be deployed due to
its accuracy level as well as its simplicity and wider adoption. It is a technique that involves

26 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

reserving a sample dataset that the model is not trained with which is then deployed on the
model at the later stages of its development before it is finalized. The steps taken in cross
validation are below:

i. Reservation of selected dataset.

ii. Training of the model with the unreserved dataset.
iii. Deployment of the validation set to ascertain the effectiveness of the model’s
performance.
The K-Fold cross validation method allows the data set to be partitioned into smaller k-
subsets and the hold-out method is applied repeatedly ‘k’ times on the subsets. The other k-1
subsets are put together as a training set each time a k-subset is utilized at a test set. The
average error from all set is later derived. The following set of steps is how a single run of k-
fold cross validation goes for any set of ‘n’ training (Ray, 2018):
1. Break datasets into k-fold in a random manner
2. Build a model on k-1 folds of the dataset for each k-fold in the dataset then
checking the model for its effectiveness for kth fold is done by testing.
3. Errors seen on each prediction is recorded.
4. Repeat 1 through 3 for all k-folds
5. The outcome is the average of the recorded errors and is called cross-validation
error. This serves as the performance metric for the model.
Movie Success System Model

The success of a movie is a combination of ratings. Input data for each movie would be a
vector of 3 dimensions containing the average score of its writers, actors and directors. The
score for each actor, writer, director is a weighted sum of the scores of all the movies each of
the dimensions have been involved in directly and indirectly. The predictor calculates the
output score of a movie from the input vector using weighted sum of k-nearest neighbours
(KNN). The weights is be optimized with cost function built by K-fold cross-validation.
The system model is designed to show an abstract representation of proposed system. That is,
a model of major system capabilities. Unified modeling language (UML) tools such as the
use case diagram and activity diagram were applied in modeling proposed system. The
following Figure 2 illustrates major interactions between the Admin and the system model:

27 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

Admin interaction with the model

Input data

Analyse data
Admin
Predict Movie
Success

Figure 2: Use case diagram for Admin interaction with the model

The role of the developer is to build an interface that allows the admin to interact with the
system. This system is designed for only the admin; there is no need to accommodate other
user roles. The admin is capable of making changes and working with the entire system. The
goal is to build a predictive model that can be used for movie success prediction and business
analysis.
Figure 3 shows the activity diagram to capture the dynamic aspect of the system. This
illustrates the dynamic behaviours of administrator’s interaction with the Movie Success
Prediction system.

Input Value
[validate]
[No]

[Yes] Error in value

Predict Model

Movie Success
outcome

Figure 3 shows the activity diagram to capture the dynamic aspect of the system

28 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

RESULT AND DISCUSSION

In the movie success prediction model, a method for reading and learning the data contained
in the IMBD dataset was implemented. Another key strategy is the choice of data mining
technique used in extracting patterns which is crucial to the prediction process. This is
important to data analysis and data evaluation. Existing dataset were analyzed to acquire
trends and patterns that could help with the prediction. This included various techniques as
data preprocessing and feature reduction which also promoted accuracy. Feature reduction is
important in overcoming the cost of dimensionality. Having ascertained the prediction
algorithm using a decision tree, the dataset were divided into train and test sets. After many
attempts, precision, accuracy and impurity measure were satisfactory. This resulted in ability
to predict the category which a movie would fall. This prediction is of great significance to
movie stakeholders and producers and with this; they would be able to take the necessary
steps to make a movie more profitable or decided against it. Also, the movie industry can use
this design to modify the movie criteria for obtaining likelihood of blockbusters. Movie
audience, in general, can also use the model to determine if a particular movie is worth it.

CONCLUSION AND RECOMMENDATION

The movie industry is one of the biggest sectors in the world. Audience acceptance of movies
is important to both participants; the customers and the business stakeholders. A movie
success prediction system capable of predicting the success of a movie is very vital especially
in a competitive market like the movie industry. This study has been able address movie
success prediction while it could be extended to tackle other related issues in the movie
industry. More so, the study is geared towards helping the movie industry and stakeholders in
reading the movie success prediction for better intelligent business decisions.

REFERENCES
Alpaydin, E. (2010). Introduction to machine learning (3rd ed.). New York: MIT Press.
Baker. R. (2010). Data Mining for Education. In McGaw, B., Peterson, P., Baker, E. (Eds.)
International Encyclopedia of Education (3rd edition), 7, 112-118. Oxford, UK:
Elsevier.
Cai, E. (2014). Machine learning of the day: The "no free lunch" theorem. [Online].
Retrieved from www.chemicalstatistician.wordpress.com/machine-learningof-the-day-
The-no-free-launch-theorem.
Dhage, S. N. & Raina C. K. (2016). A review on Machine Learning Techniques.
International Journal on Recent and Innovation Trends in Computing and
Communication, 4(3), 395-399.
Greene D., Cunningham P. & Mayer R. (2008) Unsupervised Learning and Clustering. In:
Cord M., Cunningham P. (eds) Machine Learning Techniques for Multimedia.
Cognitive Technologies. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-
540-75171-7_3
Grewal, D. S. (2014). A Critical Conceptual Analysis of Definitions of Artificial Intelligence
as Applicable to Computer Engineering. IOSR Journal of Computer Engineering
(IOSR-JCE), 16(2), 9-13.

29 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC
British Journal of Computer, Networking and Information Technology
ISSN: 2689-5315
Volume 4, Issue 2, 2021 (pp. 22-30) www.abjournals.org

Han, J., Kamber, M. & Pei, J. (2011). Data Mining: Concepts and Techniques (3rd ed.). New
York: Morgan Kaufmann.
Kumar, V., Ramakrishnan, G., & Li, Y. (2018). A framework for automatic question
generation from text using deep reinforcement learning. ArXiv, abs/1808.04961.
McCarthy, J. (2000). “Review of Artificial intelligence: A General Survey.” [Online].
Available: http://wwwformal.stanford.edu/jmc/reviews/lighthill/lighthill.html.
Mesakar, S. & Chaudhari, M. (2013). A Review of Clustering Algorithms. International
Journal of Computer Science and Technology (IJCST), 4(1), 255-257.
Nithin, V.R, Pranav, M., Badu, P. B., & Lijiya, A. (2014). Predicting movie success based on
IMDB data. International Journal of Business Intelligents, 3(2), 34-36.
10.20894/IJBI.105.003.002.004.
Ogunleye, F. (2004). A Report from the Front: The Nigerian Videofilm. Quarterly Review of
Film and Video, 21(2), 79-88. DOI: 10.1080/10509200490272991.
Ramageri, B. M. (2010). Data mining techniques and applications. Indian Journal of
Computer Science and Engineering. 1(4), 301-305.
Ray, S. (2018) Improve Your Model Performance Using Cross Validation (in Python and R).
https://www.analyticsvidhya.com/blog/2018/05/improve-model-performance-cross-
validation-in-python-r
Shu, X. (2020). Knowledge Discovery in the Social Sciences: A Data Mining Approach.
Oakland, California: University of California Press.
Usama, M., Qadir, J., Raza, A., Arif, H., Yau, K., Elkhatib, …Al-Fuqaha, A. (2019).
Unsupervised Machine Learning for Networking: Techniques, Applications and
Research Challenges. IEEE Access, 7, 65579-65615. 10.1109/ACCESS.2019.2916648.
Wales, L.M. (2017). The Complete Guide to Film and Digital Production: The People and
the Process (3rd ed.). New York: Routledge. https://doi.org/10.4324/9781315294896.
Wolpert, D. H. & Macready, W. G. (1997). "No free lunch theorems for optimization," in
IEEE Transactions on Evolutionary Computation, 1(1), 67-82. doi:
10.1109/4235.585893.

30 Article DOI: 10.52589/BJCNIT-CQOCIREC

DOI URL: https://doi.org/10.52589/BJCNIT-CQOCIREC