What You'll Learn

Predictive Modeling Course Description

Section 1: Introduction

What is Predictive Modeling

Predictive analytics is an emerging strategy across many business sectors and they are used to improve the performance of the companies. Predictive modelling is a part of predictive analytics which is used to create a statistical model to predict the future behaviour. The predictive modelling can be used on any type of event regardless of its occurrence. The predictive model to be used for a particular situation is often selected on the basis of the detection theory. This chapter includes an overview of predictive analytics and predictive modelling. This chapter also includes examples of predictive modelling.

How to Build a Predictive Model

The predictive models are used to analyze the past performance to predict the future results. There are several steps involved in building a predictive model

Pre-Processing

Data Mining

Results validation

Understand business and data

Prepare data

Model data

Evaluation

Deployment

Monitor and improve

All these steps are explained in detail under this chapter

Section 2: Variables

Types of Variables

There are different types of variables in predictive modelling. They are Predicator Variable, Numeric variable, ID variable, Factor variable, categorical variable, extraneous variable, confounding variable and Target Variable.

Difference Between Variables

The difference between each variable is explained in this chapter.

Other Types – Extraneous Variables

Extraneous variables are the ones that are not included in your test or experiment by you but it affects your experiment when you run your experiment. For example, if you are running an experiment to find out whether one variable has an effect on the other variable but unfortunately you find another variable that is affecting the result of your outcome. These undesired variables are called extraneous variables.

Section 3: Steps Included

How to Build a Predictive Model Steps

This chapter contains the basic seven steps involved in predictive modelling

Defining the objective – This section deals with the ways to define the objective of predictive models with relation to the goals of the business.

Gathering the data -Collecting data from various sources is another important step in building a predictive model. Examples are provided for the collection of different types of data from various sources.

Preparing the data for modelling – This section deals with the segregation of data and how determines how they can be used in predictive modelling.

Selecting and transforming the variables – This topic deals with the steps for the transformation of independent variables to best fit the dependent variable.

Processing and evaluating the model – In this chapter, you will go through several methods of processing and evaluating the model

Validating the model – The predictive models should perform well on the data. This chapter deals with three powerful methods for ensuring the model fit.

Implementing and maintaining the model – Effective implementation of the Predictive model is another important step. This chapter discusses various auditing procedures and model maintenance practices

Algorithms

Algorithms perform data mining and statistical analysis to find out the trends in data. The predictive analytics has few built-in algorithms like regression, time series, outliers, decision trees, k means and neural network.

Forecasting Methods

The forecasting methods used depend mainly on the type of data available. There are different methods of forecasting which are discussed in detail in this lesson

Qualitative Forecasting

Quantitative forecasting

Cross-sectional forecasting

Time Series Forecasting

What is Time Series

Time series algorithms are used to perform time-based predictions. The examples are smoothing methods which are discussed in the next chapter.

Time series data are used to forecast something that is constantly changing over time like profit, share price and others. Forecasting the time series data will help to predict the sequence of observations in the future.

Section 4: Smoothing Methods

Smoothing Methods – Moving Averages

In moving average smoothing each observation is assigned an equal weight and each observation is forecasted using the average of the previous observations. This method is useful when the item to be forecasted remains unchanged over time. The formula for moving average method is also explained in this chapter.

Smoothing Methods – Double Exponential Smoothing

Exponential smoothing is one of the successful and most widely used forecasting methods. The forecasts that are produced using exponential smoothing methods are weighted averages of the previous observations. There are two types of exponential smoothing – Simple Exponential Smoothing and Double Exponential Smoothing.

Double exponential smoothing is the recursive application of an exponential filter twice in a time series. This should not be used if the data is expected to be affected by seasonality. This is used only when a trend in inherent in the dataset. The double exponential smoothing is explained in detail in this chapter with its formula and examples.

Section 5: Regression Algorithms

Regression Algorithms – Exponential

There are different types of statistical, data mining and machine learning algorithms in Predictive Modeling. Each algorithm is used to address the specific needs of the business concern. So choosing the right algorithm for your business is a great task. Regression algorithm is one among them. Regression algorithm is used to forecast continuous data like credit scoring or predicting the next outcome of a time-based event. For example, regression algorithm can be used to predict the trend of a stock movement with its past prices.

There are seven types of regression which are explained in detail in this chapter. In predictive modelling, linear and logistic regressions are considered to be the most important ones.

Linear Regression

Logistic Regression

Polynomial Regression

Stepwise Regression

Ridge Regression

Lasso Regression

ElasticNet Regression

This chapter also lets you learn how to select the right regression model for your business.

Section 6: Clustering Algorithms

Clustering Algorithms – Definition

Clustering can be defined as the division of data into groups of similar objects. The clustering technique is not used to predict the value of the target variable in clustering. The clustering algorithm is used to segment the whole data into homogeneous clusters. For example, clustering is used for customer segmentation. This algorithm segments the customers based on more variables which can never be done by humans. To make a clustering algorithm perfect, there should be more similarity within the clusters and more differences between the clusters. There is a different type of clusters and each is explained in detail in this chapter.

Clustering Algorithms – Fuzzy C Means Clustering

Fuzzy C Means Clustering is one of the widely used clustering algorithms. This is a method of clustering which lets a single data to be present in two or more clusters. This algorithm is mostly used in pattern recognition. This algorithm is used for analysis based on the distance between various input data points. The formula for Fuzzy C Means Clustering is explained in this chapter. Here you will also learn about the algorithm steps of Fuzzy C Means Clustering. This chapter also contains the comparative study of K means Cluster algorithm and Fuzzy C Means Cluster algorithm.

Section 7: Neural Network Algorithm

Neural Network Algorithm

The neural network algorithm is used for pattern recognition, find out the predictions and learn from the result. An example of a neural network is the human brain. Neural networks in data mining apply pattern recognition and machine learning algorithms to build predictive models. This chapter explains the two main components of the neural network algorithm – Nodes and Links. Nodes are artificial neurons and Links are the components which connect these nodes. The other topics included in this chapter are

Kohonen Neural Network

Steps of algorithm of learning the neural network

Case studies

Section 8: Support Vector Machines

Support Vector Machines

Support Vector Machine (SVM) is a supervised machine learning algorithm which is used for analyzing data for classification and regression categories. But this algorithm is mostly used in classification problems. The advantages of SVM are it can be effectively used in high dimensional spaces and it is also memory efficient. The SVM is also versatile and different kernels can be used here. The main disadvantage of SVM is that it produces poor performance if the number of features is more than the number of samples. The other topics included in this section are listed below

Linear Support Vector Machines

Non Linear SVM

Basics of Support Vector Machines

Calculating the SVM classifier

How is the optimal hyperplane calculated

How to implement SVM in Python

Source code and Explanation

Tuning the parameters of SVM

Support Vector Regression

Pros and Cons of Support Vector Machines

Practice problem and Case Studies