In DataScienceSalon/mdb: Movie Rating Model and Predictor

Appendix

Appendix A: Source Code

This appendix includes the primary functions used to perform forward selection, backward elimination, and prediction. The rest of the source code is available on github at https://github.com/DataScienceSalon/mdb.

Data Preprocessing

This function performs the preprocessing of the data in advance of the analysis

Forward Selection

This function performs the forward selection process.

Backward Elimination

This function performs backward elimination.

Regression Analysis

The regression analysis function performs the tests, and renders the plots requuired to validate the regression assumptions.

Prediction

This function performs the prediction of the ten test films

Appendix B: Formulas

The cast scores and cast votes variables characterized the overall popularity of the cast. Total scores, defined as 10 * IMDb rating plus the audience score, was apportioned according to proportional allocation and this value was summed over the individual cast members. The cast score variable was derived as follows:
$$cs_i = [\displaystyle\sum_{a=1}^5 \displaystyle\sum_{f_a=1}^n s_{f_a} * p_a] - s_i $$ where:
$cs_i$ is the cast score for film $i$
$a$ is the variable containing the five credited actors
$f_a$ is the film in which actor was credited
$n$ is the number of films in which the actor was credited
$s_{f_a}$ is the total score for film $f_a$, computed as (10 * imdb_rating) + audience score
$p_a$ is the proportion of $s_{f_a}$ allocated to actor $a$ and is defined as:
Actor 1: .40 * $s_{f_a}$
Actor 1: .30 * $s_{f_a}$
Actor 1: .15 * $s_{f_a}$
Actor 1: .10 * $s_{f_a}$
Actor 1: .05 * $s_{f_a}$
$s_i$ is the total score for film $i$

The cast votes variable was computed analogously.