• Project started on November 2019
• Members
  • Katherine Fu (Junior, CS, Stat)
  • Songyang Cheng (Junior, Information systems, CS)

Details data

• Data matrix $X \in R^{n \times p}$ where n is the number of plant plots, and p are characteristics of soil in plot (chemical, microbiome)
• Response $Y$: disease of plants or yield
• Type of analysis:
  • penalized regression tools ($n < p$)
  • random forest
  • neural networks

Steps to clone the repo locally

1. In the terminal, go to the folder where you want to put the local repository using cd
2. Type:

git clone https://github.com/solislemuslab/potato-lankau.git

3. Inside this folder, create a subfolder called data where you will put the data. This folder is ignored in the .gitignore file, so nothing inside this folder will be pushed to github

Songyang's Files

• Python Scripts will be .py files
• All visualization will be .ipynb files with naming rules: week number + date
• A serious bug in work for week 3 and week 4 work has been identified in week 6. Thus, week 3 and week 4 work should not be cited. All analysis has been redone in week 6. For the sake of integrity, the buggy files are not deleted.

summary for each file

• w010119
  Conducted exploratory data analysis.
  Visualized the pattern for both potato yield and disease severity.
• w020202
  Checked y distribution. Converted to normal distribution.
  Plotted correlation Heat Map for meta data and yt.
  Plotted standard deviation for each field (biological difference).
  Tried out basic random forest model.
• w030209 (buggy)
  Plotted euclidean distance between different fields. (no bugs here)
  Tuning Random Forest models. (bugs exist in train-test split)
• w040216 (buggy)
  Explored XGBoost models. (bugs exist in train-test split)
• w050223
  Plotted correlation Heat Map for part of (selected) meta+taxa data and yt.
  Plotted linear relationship between each selected feature and y/yt.
• w060303
  Corrected bugs in week 3 and week 4 by redoing train-test split.
  Explored PCA and Keras Auto Encoder.
• w070319
  Explored Pair Wise Selection. Explored Ridge Regression.
• w080402
  Further explored Ridge Regression.
  Explored Lasso Regression.
  Visualized two best models.
• w090408
  Visualized full model selection tree.
• w100416
  Verified the difference between Linear Regression and Ridge Regression.
• w110506
  Created for the purpose of writing my report. Nothing new.