Welcome to my Machine Learning playground. SmartML is a collection of experiment machine learning algorithms that I implemented from scratch in python. The goal of this project is to get hands dirty and experience the real power of Machine Learning.

My ultimate goal is to build a ML toolbox that speed up the time on finding the best approach for your dataset. This is the reason why I call it ‘smart’ ML.

Currently, this project is very far from my ultimate goal, but it is good for anyone who want to play around the well-known ML algorithms. You are free to change or use it.

My blog also cover most of ML algorithms that I implemented in SmartML with examples, so feel free to use it as reference. Thanks.

My Blog : Make it easy

Requirement:

• Python 2.75
• Numpy
• Matplotlib
• Libsvm

(Please Go to : Libsvm to Download the library, then open a terminal direct to .libsvm.3.XX/python/ and type “make”.)

Libsvm

SmartML Algorithm include : ============

Dimension reduction:

• PCA
• Kernel PCA
• SVD
• LDA
• Factor Analysis

Normalization:

• min MAX
• Z Score
• log

Supervise Algorithms (Classifiers):

• KNN
• Logistic Regression
• ID3 Tree
• Naive Bayes
• SVM
• Adaboost

Unsupervised Algorithm :

• KMean

Regression :

• RidgeRegression
• Local weighted linear regression (lwlr)
• Regression tree (regtree)

Others:

• Hidden Markov Model

0. Import SmartML:

   from smartML.main import SmartML

# General Usage

1. first, Declare my smart ML:

   myML = SmartML()

2. load data using trainSet.txt for example:

   myML.loadfile('./trainSet.txt', ',', 'supervise', normalize=‘minMax’)

3. (Optional) plot data:

   myML.plot(1, 0)

4. (Optional) dim reduce methods call:

   myML.dimReduce('FactorAnalysis')

5. call ML methods:

Methods Include:

• unsupervise: kmeans
• regression: Linear, local weighted linear regression (lwlr) and regression tree (regtree)

• supervise: KNN, naive bayes (naivebay), logistic regression (logistic), ID3, adaboost and SVM (libsvm)

  myML.learner(method=‘ID3’)

6. test your model:

   myML.tester(test.dataset)

7. return a result label array for testing dataset

# HMM

1. construct a probability table:

   prob = {}
   
   prob['transition'] = array([[.6, .2, .1, .2], [.6, .05, .1, .25], [.15, .05, .6, .2], [.2, .1, .3, .4]])
   
   prob['emission'] = array([[.2, .1, .5, .2], [.1, .4, .3, .2], [.2, .3, .5, 0], [.2, .2, .1, .5]])
   
   prob['state'] = ['watching TV', 'Pub Night', 'Party Night', 'Study']
   
   prob['observations'] = ['tired', 'hungover', 'scared', 'fine']

2. call smartML:

   myML = SmartML()

3. call leaner method with table:

   my.learner('HMM', param=prob)

4. create a observation:

   obs = array([0, 3, 2, 3, 0, 1, 1])

5. probability return:

   my.tester(obs)

My Blog for HMM silly example

Have Fun !~