ml-exercises

Repo for our machine learning exercises in SS2020.

Authors:

- Alexander Leitner
- Aleksander Hadzhiyski
- Peter Holzner

---

The others:

- Cedrik und Co: Group 7
- Jan+Fabian: Group 33

Exercise 0: Dataset registration & description

Submission deadline: 25.03.2020, 23:59

Regression: Moneyball

Classification: Carabana - Don't get kicked

---

Exercise 1:

Description

Submission deadline: 4.05.2020, 23:59

Choose 4 techniques from:

a) linear regression
b) polynominal regression
c) logarithmic regression
d) kNN
e) Lasso
f) Ridge
g) Regression tree
h) ...?

You need to chose a total of 4 datasets.

• 1 from exercise 0
• 1 from Kaggle/UCI ML Repository (or other repositories)
• 2 from list below

Choose 2 from:

Bias correction of numerical prediction model temperature forecast Data Set http://archive.ics.uci.edu/ml/datasets/Bias+correction+of+numerical+prediction+model+temperature+forecast
Data Folder: http://archive.ics.uci.edu/ml/machine-learning-databases/00514/

Some info:

> Number of Instances: 7750
> Number of Attributes: 25
> Missing Values: Yes

Verdict:

> # of samples: med
> # of dimensions: med-high

QSAR fish toxicity Data Set http://archive.ics.uci.edu/ml/datasets/QSAR+fish+toxicity Data Folder: http://archive.ics.uci.edu/ml/machine-learning-databases/00504/

Some info:

> Number of Instances: 908
> Number of Attributes: 7

Verdict:

> # of samples: low
> # of dimensions: low

Metro Interstate Traffic Volume Data Set http://archive.ics.uci.edu/ml/datasets/Metro+Interstate+Traffic+Volume Data Folder:http://archive.ics.uci.edu/ml/machine-learning-databases/00492/

Some info:

> Number of Instances: 48204
> Number of Attributes: 9

Verdict:

> # of samples: high
> # of dimensions: low

Real estate valuation data set Data Set http://archive.ics.uci.edu/ml/datasets/Real+estate+valuation+data+set Data Folder: http://archive.ics.uci.edu/ml/machine-learning-databases/00477/

Some info:

> Number of Instances: 414
> Number of Attributes: 7

Verdict:

> # of samples: low
> # of dimensions: low

---

Meetings:

Meeting 1: Fr., 24.04.2020

Next meeting: Monday evening, 17:00

Chosen:

a) linear regression
b) kNN
c) Lasso
d) random forest --> regression tree?

How:

> Jupyter notebook
> bissl plots
> bissl code zum spielen
> bissl ergebnisse/fazits
> bissl pros/cons

Who & what (on Moneyball):

a) linear regression    --> Alex
b) kNN                  --> Aleks
c) Lasso:               --> Code von Alex
d) random forest        --> Peter

---

Dataset 1: Moneyball {low/low}

Some info:

> Number of Instances: 1230
> Number of Attributes: 15

Verdict:

> # of samples: low
> # of dimensions: low-med

---

Dataset 2: Video Games Sales (Kaggle) {med/low-med}

Dev: Peter

Video Games Sales with Metacritic ratings

Dimensionality analysis:

> Rows: 16719
> Columns: 16
> Column names: ['Name', 'Platform', 'Year_of_Release', 'Genre', 'Publisher', 'NA_Sales', 'EU_Sales', 'JP_Sales', 'Other_Sales', 'Global_Sales', 'Critic_Score', 'Critic_Count', 'User_Score', 'User_Count', 'Developer', 'Rating']

Possible targets:

> Sales: so either of ['NA_Sales', 'EU_Sales', 'JP_Sales', 'Other_Sales', 'Global_Sales']
> Rating: so either of ['Critic_Score', 'User_Score']

---

Dataset 3: Metro Interstate Traffic Volume Data Set {high/low}

http://archive.ics.uci.edu/ml/datasets/Metro+Interstate+Traffic+Volume Data Folder:http://archive.ics.uci.edu/ml/machine-learning-databases/00492/

Dev: Alex L

Some info:

> Number of Instances: 48204
> Number of Attributes: 9

Verdict:

> # of samples: high
> # of dimensions: low

---

Dataset 4: Bias correction of numerical prediction model temperature forecast Data Set {med/med}

http://archive.ics.uci.edu/ml/datasets/Bias+correction+of+numerical+prediction+model+temperature+forecast
Data Folder: http://archive.ics.uci.edu/ml/machine-learning-databases/00514/

Dev: Aleks H

Some info:

> Number of Instances: 7750
> Number of Attributes: 25

Verdict:

> # of samples: med
> # of dimensions: med-high

---

Cheat Sheets:

Free books on Data Science by Springer: https://towardsdatascience.com/springer-has-released-65-machine-learning-and-data-books-for-free-961f8181f189

Here is github repo with some cheat sheets for various Data Science/Machine Learning related cheat sheets:

https://github.com/abhat222/Data-Science--Cheat-Sheet

https://www.youtube.com/watch?v=tODN7x3BO_E three videos of linear regression

Other cheat sheets are saved in the folder of the same name.