README

Project: The influence of music

Project Member: Kaiyang Chen, Baole Fang, Yumeng Bai

Filetree

.
├── 2021_ICM_Problem_D_Data
│   ├── data_by_artist.csv
│   ├── data_by_year.csv
│   ├── full_music_data.csv
│   └── influence_data.csv
├── 2021_ICM_Problem_D_Data_std
│   ├── data_by_artist.csv
│   ├── data_by_year.csv
│   ├── full_music_data.csv
│   └── influence_data.csv
├── README.md
├── STAT4060J_Group25_Project_Final_Draft.pdf
├── step1
│   ├── influence_page_rank.py
│   ├── out_degree
│   │   ├── out_degree_0.1_0.png
│   │   ├── out_degree_0.2_0.png
│   │   ├── out_degree_0.3_0.png
│   │   ├── out_degree_0.4_0.png
│   │   ├── out_degree_0.5_0.png
│   │   ├── out_degree_0.6_0.png
│   │   ├── out_degree_0.7_0.png
│   │   ├── out_degree_0.8_0.png
│   │   └── out_degree_0.9_0.png
│   ├── page_rank_result
│   │   ├── influence_page_rank_0.05_0.csv
│   │   ├── influence_page_rank_0.1_0.csv
│   │   ├── influence_page_rank_0.2_0.csv
│   │   ├── influence_page_rank_0.3_0.csv
│   │   ├── influence_page_rank_0.4_0.csv
│   │   ├── influence_page_rank_0.5_0.csv
│   │   ├── influence_page_rank_0.6_0.csv
│   │   ├── influence_page_rank_0.7_0.csv
│   │   ├── influence_page_rank_0.8_0.csv
│   │   └── influence_page_rank_0.9_0.csv
│   ├── permutation_test_ranking.R
│   ├── sub_good.png
│   ├── subsub.png
│   └── visual.py
├── step2
│   ├── FeatureAnalysis.R
│   └── PvaluePlot.py
├── step3
│   ├── MCC.py
│   ├── original
│   │   ├── MCC.jpeg
│   │   ├── MCC_PermutationTest.jpeg
│   │   ├── artists.csv
│   │   ├── clusters.csv
│   │   └── mcc_null.csv
│   ├── similarity.py
│   ├── std
│   │   ├── MCC.jpeg
│   │   ├── MCC_PermutationTest.jpeg
│   │   ├── artists.csv
│   │   ├── clusters.csv
│   │   └── mcc_null.csv
│   └── visual.py
└── step4
    ├── SS.csv
    ├── SS.png
    ├── SS.py
    ├── SS_.csv
    ├── bootstrap.R
    └── influence_page_rank_0.5_0.csv

Step1

Use influence_page_rank.py to implement PageRank algorithm and plot relevant figures.

Then use permutation_test_ranking.R to do permutation tests on the two rankings.

Visual.py is used to generate influencial network plots.

Step2

Use python PvaluePlot.py to generate the 2-tail p-value matrix for feature pairs.

Run the code trunk in FeatureAnalysis.R step by step to get the analysis result.

Step 3: Measure of musical similarity

Overview

In this step, we purpose a measure of musical similarity. We perform a permutation test to verify its validness. The null hypothesis is that the music genres are independent of musical features.

One assumption we make is that songs/musicians within the same genre are more similar than those between different genres. If we can reject the null hypothesis, it means that our measure of musical similarity is valid.

Data

• data_by_artist.csv
• influence_data.csv

How to run

1. Run MCC.py and get artists.csv, clusters.csv

To run scripts successfully:

• Please specify the path of input and output
• Please install the necessary packages by yourself

Results

Step 4: Influence and similarity

Overview

In this step, we analyze the relationship between musical influence and similarity. We perform a stratified bootstrap to find the confidence interval of the difference of SS mean. If the confidence interval lies above 0, it means that high-influential musicians have a stronger impact on their followers.

Data

• influence_page_rank_0.5_0.csv

How to run

1. Run SS.py and get SS.csv
2. Replace "0.0" in SS.csv with "nan" (data cleaning)
3. Run bootstrap.R to compute the confidence interval

To run scripts successfully:

• Please specify the path of input and output
• Please install the necessary packages by yourself

Results