2017 - Predicting Oscar winners & box office hits

Description

Using Jupyter notebooks and scikit-learn, you’ll predict whether a movie is likely to win an Oscar or be a box office hit. Together, we’ll step through the creation of an effective dataset: asking a question your data can answer, writing a web scraper, and answering those questions using nothing but Python libraries and data from the Internet.

Abstract

Using Jupyter notebooks and scikit-learn, you’ll predict whether a movie is likely to win an Oscar or be a box office hit. Together, we’ll step through the creation of an effective dataset: asking a question your data can answer, writing a web scraper, and answering those questions using nothing but Python libraries and data from the Internet.

This talk is for engineers, data scientists, and movie lovers who want to learn how to scrape information from the Internet, and then use python libraries (and some domain knowledge) to answer interesting questions using that data. This presentation could be informative for people with a wide range of skill-levels, but I expect it to be especially useful for anyone getting started with data science, http requests, pandas, and sklearn.

By the end of this talk, the you should expect to (a) understand how to scrape and manage small to medium data sets, (b) know how to overcome the most common roadblocks (i.e. dealing with timeouts or API keys), (c) understand the tools you need to use and steps you need to take to answer interesting questions in data science, and (d) have access to a great example project in a Jupyter notebook that you can use as a template or extend.

Bio

Deborah is a PhD student, studying machine learning at Harvard University, and she graduated from MIT with a M.Eng. in Electrical Engineering & Computer Science. Her work in machine learning has spanned developing models of human perception to exploring medical data. She has also been awarded the NSF, Fulbright, and ACM/Intel Computational & Data Science Fellowship. She has spoken at PyTennessee, SciPy Conf, AI With the Best, QConNY, and PyCon US.

https://www.youtube.com/watch?v=5R1OwnGjRQM

Transcript

1. Predicting box office hits & Oscar winners using things you found on the Internet Deborah Hanus @deborahhanus
2. Background
3. Credit: Gil Press, Forbes 82% 13% 5%
4. 82% DATA WRANGLING 13% EDA & ML 5% OTHER 62% DATA WRANGLING 20% DATA UNDERSTANDING
5. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
6. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
7. What factors drive movie revenue? Image: unclaimedmoney.com
8. Not so good: Vague Will my movie be a box office hit? Good: Likelihood What is the likelihood that my movie will be a box office hit given that it has X features? Good: Correlation What attributes of a movie are correlated with box office success? Define a question you can answer
9. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
10. • Relevant • Structured • (Relatively) complete What is good data?
11. Where to find good data?
12. • Use an API • Write a web scraper • Get all the text • Make the text queryable How to get good data
13. • Make an HTTP request to get the HTML Writing a web scraper Requests Example
15. Writing a web scraper http://www.boxofficemojo.com/yearly/chart/? page=1&view=releasedate&view2=domestic&yr=%2 017.htm
18. • Make HTML queryable using BeautifulSoup & PyQuery Writing a web scraper
20. Right Click : View Source
26. • Rate limiting • API Keys • Selenium Common problems
27. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
28. Factors we can explore: • Movie budget • IMDB Rating • Power Studios • Opening Weekend • How many opening theaters • Seasonality • MPAA Rating Exploratory Data Analysis
29. Gross revenue vs. # opening theaters Exploratory Data Analysis ~3500
30. Gross revenue vs. Quality rating Exploratory Data Analysis No relationship
31. Gross revenue vs. Opening gross Exploratory Data Analysis Predictive
32. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
33. Multivariate regression Exploratory Data Analysis Predictive
34. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
35. • Budget helps (but only a little). • Timing is important. December is a great release date. • PG & G rated movies make more. • Money made in opening weekend is important. What did we find?
36. What does it take to win an Oscar?
37. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
38. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
39. What makes an Oscar winner? Image: superawesomevectors.com
40. Not so good: Vague Will this movie win an Oscar? Good: Likelihood What is the likelihood that this movie will win an Oscar given that it has X features? Good: Correlation What attributes of a movie are correlated with the movie winning an Oscar? Define a question you can answer
41. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
42. Acquiring Data IMDBpy Drama
43. Not so good: Vague Will this movie win an Oscar? Good: Likelihood What is the likelihood that this movie will be a box office hit given that it has X features? Good: Correlation What attributes of a movie are correlated with the movie winning an Oscar? Define a question you can answer Best: Conditional correlation Given that a movie has been nominated for an Oscar, what attributes are correlated with winning?
44. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
45. Factors we can explore: • Movie nomination category • Thematic content (e.g. family, violence, war, father- son relationship, smoking) • Movie genre • Where the movie was made • When the movie debuted Exploratory Data Analysis
46. Exploratory Data Analysis Countries associated with winning Oscar movies
47. Exploratory Data Analysis Number of winning movies per month
48. Exploratory Data Analysis Ratio of winning films to films to nominated films by month
49. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
50. • Binary output - winner/non-winner • More accurate than baseline What do we want from a model?
51. • Binary output - winner/non-winner • More accurate than baseline What do we want from a model?
52. • Logistic regression (Ridge, Lasso, ElasticNet) • Support vector machine (SVM) • Ensemble methods Potential models
53. • Binary output - winner/non-winner • More accurate than baseline What do we want from a model?
54. • All winners Accuracy = 29% • All losers Accuracy = 71% Establish Baselines
55. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
56. To select a model, think about what it gets wrong.
57. Selecting a model Confusion Matrix
58. • Accuracy = (TP+TN)/(TP+TN+FN+FP) • Recall = TP/(TP+FN) • Precision = TP/(TP+FP) • F1 = (Precision*Recall)/(Precision+Recall) Selecting a model
59. Selecting a model Receiver-Operating Characteristic (ROC)
60. • Define a question you can answer • Acquire good data • Understand data • Fit a model & analyze error • Draw conclusions from the data How to build & use a great dataset
61. • Nominated films made in Italy & Spain have a good chance at winning. • Films are more likely to win if they are released later in the year. • Tone down the gore (unless it is a war film). • If a film is nominated for best picture, its odds of winning are good. • If a film is nominated for best cinematography, its odds are less good. What did we find?
62. What did our model get wrong?
63. • What did your model misclassify? • Are any of those errors systematic? Analyze errors
64. Image: NYT Coded Gaze: Joy Buolamwini
65. Labeling sensitive content
66. Always analyze your errors
67. • Defined an answerable question • Built a web scraper • Explored the data • Fit a model to the data • Analyzed our errors What did we do?
68. Team
69. Building a scraper Requests - HTTP for Humans BeautifulSoup or PyQuery Analyzing data Jupyter SciKit Learn Statsmodels Example Projects http://oscarpredictor.github.io Where to go from here? Deborah Hanus @deborahhanus