Python tutorial for ML

©2018 Extreme Networks, Inc. All rights reserved
Bin Han
Serial GTAC ENG in APAC
Python Tutorial for Machine Learning

Tutorial 1: Check Version
2

Tutorial 2: NumPy, Pandas and Matplotlib
3
print myfirstarray
print myfristdataframe
plt.show()

Tutorial 3: Load CSV Data from Pandas
4
print data.head(10)
‘https://goo.gl/bDdBiA’
is Pima Indians Diabetes
from UCI Machine Learning Repository

Tutorial 4: Data Description in Pandas
5

Tutorial 5: Visualize Data
6

Tutorial 6: Pre-Processing Data for Modeling
7
See Definition of Pre-processing Data

Tutorial 7: Resampling Methods
8
Accuracy =
!"#$$%"&
!&#&'(
See Definition of k-fold Cross Validation

Tutorial 8: Algorithm Evaluation Metric for Classification Model
9
The following Evaluation Metrics are
used to classification model:
• Accuracy
• Precision
• Recall
• F score
• ROC
• AUC
• Log Loss
See Definition of Log Loss

Tutorial 9: Algorithm Evaluation Metric for Regression Model
10
The following Evaluation Metrics are used to
regression model:
• R2 (Coefficient of determination)
• MAE(Mean Absolute Error)
• RMSE(Root Means Squared Error)
• MAPE(Mean Absolute % Error)
• MSE(Mean Squared Error)
See Definition of MSE
‘https://goo.gl/FmJUSM’
is Boston House Price dataset
from UCI Machine Learning Repository

Tutorial 10: Algorithm Comparison
11
See Machine Learning Algorithm

Tutorial 11: Improve Accuracy with Tuning the Hyperparameter
12
See Tuning the Hyperparameter

Tutorial 12: Improve Accuracy with Ensemble Learning
13
See Ensemble Learning

Tutorial 13: Finalize and Save the Model
14
Save the serialized format to a file
Load the model and evaluate it
See Model Validation Strategy

Summary: Machine Learning Workflow
15
Tutorial 3, 4, 5
Tutorial 6
Tutorial 7, 8, 9
Tutorial 10
Tutorial 11, 12
Tutorial 13

Reference
16
Tutorial 1 Installing scipy
Installing numpy
Installing matplotlib
Installing pandas
Installing sklearn
Tutorial 2 random
cumsum
Tutorial 3 pandas read_csv
Tutorial 4 pandas describe
Tutorial 5 pandas values
Tutorial 6 pandas values
scikit-learn preprocessing data
scikit-learn standard scaler
Tutorial 7 model_selection KFold
model_selection cross_val_score
Tutorial 8 scoring parameter
Tutorial 9 mean squared error
Tutorial 10 supervised learning
unsupervised_learning
Tutorial 11 linear_model RidgeCV
model_selection GridSearchCV
Tuning the hyper-parameters
Tutorial 12 ensemble VotingClassifier
Tutorial 13 pickle operation
Data Sets Model UCI Machine Learning Repository
Workflow Machine Learning in MATLAB

Knowledge Reference

Tutorial 6: Math Definition of Pre-Process Data
Data normalization re-scales numerical values to a specified
range. Methods include:
§ Min-Max Normalization
– Linearly transform the data to range, i.e., 0 ~ 1, where mix value
is “0” and max value is “1”
– !"#$% =
'(')*+
'),-(')*+
§ Standardization
– Scale data based on mean and standard deviation
– z =
' ( .
/
§ Decimal scaling
– Scale the day by moving the decimal point of the attribute value
18
CLICK HERE
To GO BACK

Tutorial 7: Resampling Methods for k-fold cross validation
K-fold cross validation: the data set is divided into k equal size subset
The data set is divided into ”k” subsets, and the hold-out validation is repeated ”k” times.
19
CLICK HERE
To GO BACK

Tutorial 8: Algorithm Evaluation Metric for Log Loss
Logistic Regression Loss function (Log Loss)
The loss function for linear regression is square loss. The loss function for logistic regression is Log Loss.
20
Log Loss vs. Accuracy
§ Accuracy is the count of predictions where your predicted value equals the actual value.
§ Log Loss takes into account the uncertainty of your prediction based on how much it varies from the
actual label
CLICK HERE
To GO BACK

Tutorial 9: Algorithm Evaluation Metric for MSE
21
Mean Square Error (MSE)
• MSE measures the average sum of
squares of the difference between
the actual value the predicted values
for all data points. Because of the
square, the negative values do not
cancel positive values and it also
amplifies the impact of the errors
• Because of the square, large errors
have relatively greater influence on
MSE than do the smaller error
• The best regression is the one the
minimizes MSE. Thus, a smaller
score is better
CLICK HERE
To GO BACK

Tutorial 10: See Some of Machine Learning Algorithms
22
CLICK HERE
To GO BACK

Tutorial 11: Tuning the Hyperparameter
23
Hyper-parameter tuning is an iterative process
You can tune hyperparameters manually. You can
explore many combinations of hyperparameter
values until you find a great combination. But this
would be very tedious work. Instead you can
better approaches by using
l Grid Search
l Random Search
l Bayesian Optimization
Grid search tries the exhaustive searches for all the
possible hyperparameter combinations.
CLICK HERE
To GO BACK

Tutorial 12: Ensemble Learning
24
Ensemble learning is a machine learning paradigm where multiple learners are trained can combined to solve the same
problem. By using multiple learners, the generalization ability of a ensemble can be much better than that of a single learner.
Wisdom of the crowd
CLICK HERE
To GO BACK

Tutorial 13: Model Validation Strategy
25
CLICK HERE
To GO BACK

WWW.EXTREMENETWORKS.COM

Python tutorial for ML

More Related Content

What's hot (20)

Similar to Python tutorial for ML (20)

More from Bin Han (13)

Recently uploaded (20)

Python tutorial for ML