Gain insights into data science fundamentals, explore machine learning and big data, and delve into real-time projects. Discover systematic approaches to data acquisition, wrangling, and solving diverse problems.

new.tar.gz

data_science_notebooks

docker_env

data_science_notebooks-copy

There is a lot of dispersed, and somewhat conflicting information on the internet when it comes to data science, making it tough to know where to start. Don't worry. 

This course will get you familiar with the state of data science and the related fields such as machine learning and big data. You will be going through the fundamental concepts and libraries which are essential to solve any problem in this field. 

You will work on real-time projects from Kaggle while also honing your mathematical skills which will be used extensively in most problems you face. You will also be taken through a systematic approach to learning about data acquisition to data wrangling and everything in between. This is your all-in-one guide to becoming a confident data scientist.

An Introductory Guide to Data Science and Machine Learning

# Decision Trees

Decision Trees are powerful and provide an output that domain experts and practitioners can easily understand. Decision Trees provide the basis for many **Ensemble Methods**, which involve using multiple models for inference and producing the output for the datasets at hand.

Decision Tree, as the name suggests, is constructed in a Tree manner, including a **root node**, **internal nodes**, and **leaf nodes**. Leaf nodes, also known as terminal nodes, give us the class of the instances falling in that terminal node, and the goal is to have homogeneous terminal nodes. Root Node refers to all the instances in the dataset. Interior nodes partition the set of instances. Once created, a tree can be navigated with a new row of data following each branch with the splits until a final prediction is made.





The above Decision Tree distinguishes between males and females.  The Decision Tree displayed above can be represented in the form of **if statements** as seen below.
`````
    If Height > 180 cm Then Male
    If Height <= 180 cm AND Weight > 80 kg Then Male
    If Height <= 180 cm AND Weight <= 80 kg Then Female
```````   
## Decision Tree algorithms 





### ID3
ID3 stands for the Iterative Dichotomiser 3 algorithm. It was developed by Ross Quinlan in 1986, and it was a predecessor of algorithms like C4.5. The algorithm works by finding categorical features for each node in the tree, which gives us the maximum information gain for the categorical targets. Trees grow to their maximum size, exhausting all the features. We perform the pruning step in Decision Trees to help the Trees generalize well on the unseen dataset.


### C4.5

The C4.5 algorithm is a successor to the ID3 algorithm.  It removed the restriction that features should be categorical to build the tree, unlike the ID3 algorithm. 
This algorithm is used by partitioning the continuous or numerical input features into a discrete set of intervals. It converts the trained trees into sets of if-then rules.   

### C5.0

C5.0 is the latest version released under a proprietary license. It uses less memory and builds smaller rulesets than C4.5 while being more accurate.


### CART 


CART is an acronym for Classification and Regression Trees. It is very similar to the C4.5 algorithm. However, it provides an additional advantage because it also deals with Regression Problems and does not compute rule sets. It builds the Decision Tree using the features that yield the maximum information gain at each node.
 


# CART (Classification and Regression Trees)

It constructs a Binary Tree.  Constructing the Decision Tree involves a number of steps. These steps are divided into **induction** (training the model) and **pruning** of the constructed tree. Pruning is the step of removing an unnecessary structure from the Decision Tree to avoid overfitting. 

## Induction and pruning

From a high level, decision tree induction goes through 4 main steps to build the tree:

1. It begins with a training dataset. It has input features and target features, which may be discrete for classification problems or continuous for Regression problems.
2. We pick the best feature to use as criteria to split the dataset. Picking the best feature involves some steps, which we will look into in the next lesson.
3. We split the dataset into subsets that contain the possible values for the best feature chosen. This process of splitting introduces a node in the Decision Tree. Each node is formed based on one of the dataset’s features.
4. In this way, we recursively generate the new tree nodes by using a subset of the data created from step 3. After the Decision Tree has been built, we can use the below-mentioned techniques to deal with overfitting and to make the best Decision Tree models:

**Pre-Pruning:** We can stop making a Decision Tree if we achieve a lower bound for the number of instances in the nodes or if we achieve a lower bound for the number of instances.

**Post-Pruning:** We can create the whole Decision Tree first and then prune the branches, which do not add much. The common approach includes splitting the dataset into training and validation dataset and monitoring the performance while building the model.





Step 2 involves mathematical detail. Selecting the best feature to split on, is a **Greedy Procedure**. 


* For **Classification** problems, we use the **Gini-Index** function. It is stated below:




---
$$\sum (p_k * (1-p_k))$$
---


$p_k$ is the proportion of training instances of class $k$ in a particular prediction node. An error value of zero is ideal for a node, which means that each split will output a single class 100% of the time. Once we get to that particular decision node, we know exactly what our output will be. We also know whether we are on one side of the decision boundary or the other. The concept of having a single class per split across our dataset is called **information gain**. We will maximize information gain. Check out the example below.

Decision Trees algorithms are versatile and easy to understand models in Machine Learning. It makes a model by learning decision rules from the underlying dataset. You will learn more in this lesson. 

What is Data Science ?

Applications of Data Science

Overview of Libraries

Probability and Statistics

Machine Learning Part-1

Machine Learning Part-2

Machine Learning Part-3

Deep Learning

Machine Learning Tools and Libraries

Big Data Tools and Technologies

Where to go next ?

Decision Trees

Decision Trees