Data-Driven Blockbuster Planning on Online Movie Knowledge Library

Data-Driven Blockbuster Planning on Online Movie Knowledge Library

Data-driven Blockbuster Planning on Online Movie Knowledge Library Ye Liu∗, Jiawei Zhangy, Chenwei Zhang∗, Philip S. Yu∗z ∗Department of Computer Science, University of Illinois at Chicago, IL, USA fyliu279, czhang99, [email protected] yDepartment of Computer Science, Florida State University, FL, USA [email protected] zInstitute for Data Science, Tsinghua University, Beijing, China Abstract—In the era of big data, logistic planning can be serve as a movie knowledge library to help achieve better made data-driven to take advantage of accumulated knowledge results for future movie planning. Therefore it is no longer in the past. While in the movie industry, movie planning can efficient to rely on conventional heuristics for comprehensive also exploit the existing online movie knowledge library to achieve better results. However, it is ineffective to solely rely movie planning [1]. Data-driven movie planning methods are on conventional heuristics for movie planning, due to a large in great need to exploit the accumulated knowledge to support number of existing movies and various real-world factors that the decision-making process when planning for a new movie. contribute to the success of each movie, such as the movie The data-driven planning has shown a huge success on the genre, available budget, production team (involving actor, actress, well-known TV series “House of Cards”, produced by Netflix, director, and writer), etc. In this paper, we study a “Blockbuster 1 Planning” (BP) problem to learn from previous movies and using the data collected from viewer . plan for low budget yet high return new movies in a totally Generally, popular movie genres and renowned movie stars data-driven fashion. After a thorough investigation of an online are the favorable choices during the planning so as to max- movie knowledge library, a novel movie planning framework imize the gross. But remuneration of the movie stars and “Blockbuster Planning with Maximized Movie Configuration movie’s available budget also need to be considered in the Acquaintance” (BigMovie) is introduced in this paper. From the investment perspective, BigMovie maximizes the estimated movie planning. Meanwhile, a seamless collaboration among gross of the planned movies with a given budget. It is able to team members is the premise of high gross. For example, it accurately estimate the movie gross with a 0:26 mean absolute will always be easier for directors to continue working on a percentage error (and 0:16 for budget). Meanwhile, from the new movie with the movie genre and production team mem- production team’s perspective, BigMovie is able to formulate an bers that they are acquainted with. And the old acquaintances optimized team with people/movie genres that team members are acquainted with. Historical collaboration records are utilized can always have a tacit understanding and easy to arouse spark to estimate acquaintance scores of movie configuration factors when they cooperate in their new movies. via an acquaintance tensor. We formulate the BP problem as Problem Studied: In this paper, a research problem, namely a non-linear binary programming problem and prove its NP- the “Blockbuster Planning” (BP) problem, is introduced. Given hardness. To solve it in polynomial time, BigMovie relaxes the an online movie knowledge library which consists of the hard binary constraints and addresses the BP problem as a cubic programming problem. Extensive experiments conducted existing movie information, we plan the movie configuration on IMDB movie database demonstrate the capability of BigMovie including genre and production team for a new movie under a for an effective data-driven blockbuster planning. pre-specified budget. We note that although there are occasions Index Terms—Knowledge Base Discovery; Blockbuster Con- where a low budget production with unknown stars becomes a figuration Planning; Data-driven Application hit, we focus on the common cases involving known persons arXiv:1810.10175v1 [cs.AI] 24 Oct 2018 with available data. The objective of an optimal planning is I. INTRODUCTION to achieve: (1) the maximized gross, and (2) the optimized The movie industry attracts great interests from both movie acquaintance among the movie configuration factors. investors and the public audience because of its high profits The BP problem studied in this paper is a novel research and entertainment nature. Attracted by the huge market, lots problem, and few existing methods can be applied to solve of investors are inquiring about identifying high-gross movies it. The BP problem significantly differs from related works, to invest in. Besides recognizing profitable movies, it is such as (1) movie gross prediction [2], (2) viral marketing rewarding to provide a reasonable and promising planning for [3], [4], (3) team formation [5], [6]. (1) The movie gross a new movie at its developmental stage, which has been greatly prediction problem [7] studied in existing works merely fo- ignored in previous works due to the complexity of various cuses on inferring the movie gross while the BP problem factors, including the movie genre and production team (actor, aims at providing the optimal planning of various movie actress, writer and director). factors which can lead to the optimal gross for investors. (2) The booming movie industry has accumulated thousands of previous movies as well as their gross statistics, which may 1https://thenextweb.com/insider/2016/03/20/data-inspires-creativity/ BP and the viral marketing problems [4] are both planning the hard constraints, we introduce an approximation solution to problems that aimed at maximizing certain target objectives, resolve the problem in polynomial time. For the experimental but they are solving totally different problems in distinct result, we can see BigMovie outperforms the competitors. In scenarios: a) viral marketing problems are usually studied in addition, at the end of the paper, the case study is provided, online social networks based on certain information diffusion which demonstrate that by using BigMovie, a lucrative movie models, while the BP problem is studied in the online movie planning can be achieved. knowledge libraries instead; b) viral marketing problems aim at maximizing the infected users, while BP’s objective is to II. PROBLEM FORMULATION maximize the movie gross; c) instead of selecting the optimal In this section, we will first define several important con- users in viral marketing problems, the BP problem aims at cepts used in this paper, and then provide the formulation of planning for an optimal movie factor configurations. Recently, the BP problem. a variation of the LT model named PNP [8] is proposed for the movie design problem. The objective of PNP is very similar A. Notation to our work except that PNP aims to attract most of the At the beginning of this section, we will first define some target users but our model aims to achieve the maximum gross notations used in this paper. Throughout this paper, we will under the given budget. (3) Different from conventional team use lower case letters (e.g., x) to denote scalars, lower case formation problems [5], where team members are planned bold letters (e.g., x) to denote column vectors, upper case for the entrepreneurial team project base on satisfying skill letters (e.g., X) to denote elements of matrices, upper case qualification and minimizing the communication cost of the calligraphic letters (e.g., X ) to denote sets, and bold-face team members, our method also aims to maximize movie upper case letters (e.g., X) to denote matrix and high-order gross. tensors. T is used to represent the transpose of a vector (e.g., T The BP problem is challenging to solve due to: x ). jj · jj1 is the `1-norm of vector (e.g., jjxjj1). • Unknown Movie Success Factors : What are the contribut- B. Terminology Definition ing factors in the success of a movie? Few research works have ever been studied this problem, and relevant movie Definition 1. Online Movie Knowledge Library: An on- factors are still unknown. line movie knowledge library can be represented as an • Movie Gross/Budget Function: How much gross (budget) undirected graph G = (M, C, E, A), where node set can a movie make (require), given a configuration of the M = fm1; m2; :::; mng denotes the set of n movies movie success factors? A proper estimation of the movie in the library and C = fc1; c2; :::; clg is the set of l gross and budget will be required for studying the BP production team members. The node set C can be divided t s w d problem. into C [C [C [C , which denote the set of actors, • Movie Configuration Acquaintance Function: How to actresses, writers and directors, respectively. Link E represents compute the acquaintance scores among the movie con- the relationship between movie production team and movies. figuration factors? A function that can measure acquain- For instance, link ((ci; mj) 2 E) indicates participation of a tance properly is needed in defining the BP problem. production team member ci in a movie mj. And set A denotes • NP Hardness: Based on our analysis, we demonstrate that the attribute of node set M. For the movie mi, the relative attribute is A = Ag [ fab ; ag g , where Ag is the BP problem is actually an NP-hard problem, and no (mi) (mi) (mi) (mi) (mi) b solution exists that can solve it in polynomial time if P the genre list of movie mi, a(m ) is the budget of movie mi g i 6= NP. and a is the gross of movie mi. (mi) To solve the aforementioned challenges, a new movie planning Definition 2. Movie Configuration: Each movie mi 2 M in framework “Blockbuster Planning with Maximized Movie the online knowledge library will have an unique configuration Configuration Acquaintance” (BigMovie) is proposed in this , covering movie production team (involving actor, actress, paper.

View Full Text

Details

  • File Type
    pdf
  • Upload Time
    -
  • Content Languages
    English
  • Upload User
    Anonymous/Not logged-in
  • File Pages
    11 Page
  • File Size
    -

Download

Channel Download Status
Express Download Enable

Copyright

We respect the copyrights and intellectual property rights of all users. All uploaded documents are either original works of the uploader or authorized works of the rightful owners.

  • Not to be reproduced or distributed without explicit permission.
  • Not used for commercial purposes outside of approved use cases.
  • Not used to infringe on the rights of the original creators.
  • If you believe any content infringes your copyright, please contact us immediately.

Support

For help with questions, suggestions, or problems, please contact us