Description
1. Setting up an Integrated Development Environment (IDE)
There are many ways to setup the IDE. However, we recommended using Anaconda,
which is a pre-built python environment with bundles of useful packages.
(a) Download Anaconda from: https://www.anaconda.com/distribution/.
Chooose the correct version based on your operating system and install it step by
step.
(b) Configure the PATH environment variable to make the conda command work.
The following command is an easy way for testing whether your configuration is
correct. If it is, you will see the version of conda being outputed. Otherwise, for
macOS or Linux, the /path/to/anaconda/bin should be appended to PATH, and
for Windows, /path/to/anaconda, /path/to/anaconda/Library/mingw-w64/bin,
/path/to/anaconda/Library/usr/bin, /path/to/anaconda/Library/bin,
/path/to/anaconda/Scripts should be appended to PATH. 1
command: conda –version
sample output: conda 4.6.12
2. Google Colab
Google Colab is a great resource for any Machine Learning class. Follow the steps
in https://colab.research.google.com/notebooks/intro.ipynb to become more
familiar with Google Colab.
3. Pandas
This section is for practicing basic funtions of the Pandas library using the Salaries
data set.
(a) Consider the Salaries.csv file.
(b) Use the read_csv(…) method from Pandas (Documentation Link) to read data
from file Salaries.csv and to copy it into a dataframe.
(c) Make the column playerID in the csv file as the index column and the first row
as the header. Also, skip the second row when reading the file.
(d) Select the id of the players who are registered in ATL and HOU whose salary is
higher than one million.
(e) Use the describe() method to calculate the standard deviation, first quartile, median, third quartile, mean, maximum, and minimum of the salary in team ATL.
(f) Create a Python dictionary object whose keys are the headers of the dataframe
created in the read_csv() exercise and values are Python list objects that contain
data corresponding to the headers. (Here, use the iterrows() method to iterate
each row of the dataframe and copy it to a dictionary. However, there is an easier
way. Learn how the to_dict() method works by yourself later)
1
If you have no idea with what the PATH environment variable is, please use the resources on the Internet
to learn about it.
1
Homework 0 DSCI 552, Instructor: Mohammad Reza Rajati
(g) Create a dataframe using pd.DataFrameRead (Documentation Link) and from
the dictionary created in (e). Then, change the header to “a”, “b”, “c”, … .
4. Numpy
Quick start:
https://www.numpy.org/devdocs/user/quickstart.html
Numpy axes explaination:
(a) Create a 2-dimensional Python list object, then convert it to a Numpy array
object.
(b) Examine the ndim, shape, size, dtype, itemsize, and data attributes of the numpy
array object. Make sure you understand their functions.
(c) Learn the dimension concept of an ndarray object by using reshape() and flatten()
methods.
(d) Understand how the slice operation works for 1-D arrays and 2-D arrays and
practice by yourself.
(e) Learn operations on ndarray by examining the argmin(), argmax(), min(), max(),
mean(), sum(), std(), dot(), square(), sqrt(), abs(). exp(), sign(), and mod()
methods. Make yourself comfortable with these methods.
(f) Examine the arange(), ones(), zeros(), eye(), linspace(), and concatenate() methods. Make yourself comfortable with these methods.
5. Scikit-Learn
This section introduces some packages (or methods) in Python (Scikit-Learn and Scipy)
that will be frequently used in your programming assignments. You must become
familiar with them and use them masterfully.
• Data Preprocessing (Documentation Link)
– Standardization: StandardScaler
– Normalization: MinMaxScaler
– Quantifying Categorical Features: LabelEncoder. OneHotEncoder
– Construct Train and Test Sets: model_selection.train_test_split
• KNN: KNeighborsClassifier
• Linear Regression: LinearRegression
• Logistic Regression: LogisticRegression, LogisticRegressionCV
• Feature Selection / Model Selection
– L1 Penalized Regression (Lasso Regression) with Cross-Validation: LassoCV
– L2 Penalized Regression (Ridge Regression) with Cross-Validation: RidgeCV
2
Homework 0 DSCI 552, Instructor: Mohammad Reza Rajati
– Cross-Validation: StratifiedKFold, RepeatedKFold, LeaveOneOut, KFold,
model_selection.cross_validate, model_selection.cross_val_predict,
model_selection.cross_val_score
– Model Metrics (Documentation Link): accuracy_score, auc, f1_score,
hamming_loss, precision_score, recall_score, roc_auc_score
• Decision Tree: DecisionTreeClassifier, DecisionTreeRegressor
• Bootstrap, Ensemble Methods
– Bootstrap: bootstrapped (Documentation Link)
– Bagging: RandomForestClassifier, RandomForestRegressor
– Boosting: AdaBoostClassifier, AdaBoostRegressor
• Support Vector Machines (Documentation Link): LinearSVC, LinearSVR
• Multiclass and Multilabel Classification (Documentation Link)
– One-vs-one Multiclass Strategy: OneVsOneClassifier
– One-vs-the-rest (OvR) multiclass/multilabel strategy / OneVsRestClassifier
• Unsupervised Learning
– K-means Clustering: KMeans
– Hierarchical Clustering: scipy.cluster.hierarchy (not scikit-learn)
• Semi-supervised Learning (Documentation Link)
6. Git and GitHub
(a) In the directory of this jupyter notebook file locates, initiate a Git repository.
(b) Check out a new branch called dev and commit the current notebook within this
branch.
(c) Merge the dev branch to the master branch (the default branch).
(d) Create a temporary repository (just for practicing and you can delete it later) in
GitHub.
(e) Push new changes in the master branch to the remote repository created in step
(d).
(f) Checkout the dev branch again and do some changes to your notebook, and then
repeat step (c) and step (e).
7. Matplotlib
Quick start:https://matplotlib.org/3.1.1/tutorials/introductory/pyplot.html
(a) Create two one dimensional arrays x and y and plot y vs x, add title, xlabel,
ylabel, grid.
(b) Create multiple arrays and plot them with different styles, add legends, add
text/mathematical equations on the plot.
3
Homework 0 DSCI 552, Instructor: Mohammad Reza Rajati
(c) Create multiple subplots, play around with the figure size, text font/size.
(d) Get familiar with get current axis (gca) handle to do the above tasks
(e) Change the limits on x and y axes, use logarithmic axes to plot.
8. Seaborn
Quick start:https://seaborn.pydata.org/introduction.html
(a) Use the Salaries.csv file in Pandas section.
(b) Create a dataframe and try to plot it with seaborn.
(c) Perform statistical estimation on the data using seaborn in-built functions – lmplot, catplot, relpolt.
(d) Create axis level functions like boxplot to visualize
(e) Visualize the dataset structure using pairplot and jointplot.
4
