ENGG2851 Project Analytics Week 9: Predictive Analytics Presented by: Dr. Mehdi Rajabi Asadabadi ENGG2851 Course Modules Weeks Topic Week 1 Introduction and Overview Week 2 Data Basics Week 3 Basic Analytical tools Week 4 Descriptive Analytics Week 5 Diagnostic Analytics Week 6,7 Data Visualisation and Dashboarding Week 7,9 Predictive Analytics Week 8 Online Quiz Week 9 Prescriptive Analytics Week 10 Automated tools Week 11 Student presentations Week 12 Student Presentations Week 13 Online Quiz Predictive Analytics Predictive analytics focus on using historical data to identify patterns enabling the prediction of future events. The identified pattern or trend from historical data is represented by a mathematical model. This model can then be used to predict future events based on the previous data and the new data. Descriptive Analytics (Analysis) Diagnostic Analytics Predictive Analytics Prescriptive Analytics Complexity and Cost A d d e d V a lu e Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data, based on the existing data. Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data, based on the existing data. Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data, based on the existing data. Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Binary classification: Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Binary classification: Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Multiclass classification: Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Decision Tree A decision tree is essentially a tree shaped diagram used to classify. It is a predictive model based on a branching series of Boolean tests. It has a root node which is the starting point of the decision tree. Splitting or branching is the process of dividing a node into two or more sub-nodes. Decision nodes are any node that has sub-nodes and leaf/terminal nodes are the ones without a split. Sub-nodes of a specific node is known as child nodes and the node is known as parent node. Decision Tree Decision Tree Decision Tree In small datasets with limited attributes, it is possible to do the graphing of decision tree by hand, in larger datasets, it may become really time consuming and confusing. Later in your jobs, if you wanted to use decision tree, remember that there are automated tools to do decision tree for you. Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Decision trees Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Decision trees Random forest Predictive Analytics Models Classification: Classification is the process of creating a set of classes for data based on the existing data. Decision trees Random forest Voting classifiers Classification: Classification is the process of creating a set of classes for data based on the existing data. Decision trees Random forest Voting classifiers Neural networks and deep learning … Predictive Analytics Models Predictiv
e Analytics Models Classification Classification is a mathematical model that can differentiate between two or more outcomes. For instance, given the atmospheric data for the last two weeks, the decision of ‘will it rain ’ can be answered by a classification model. Instances like this where outcome is known as binary classification. A popular classification technique is decision trees. Regression As opposed to predicting a decision, regression focus on predicting a number using available data. This is typically used in banking, investing and other finance models. For instance, given the bitcoin prices for the last year, ‘what would be the price today ’ can be answered by a regression model. Other models (Clustering, Time Series, Forecasting, and similar) Regression Regression investigates the relationship between a dependent (target) variable and independent variables (predictor) It is used for forecasting, time series analysis and finding causal effect relationship between variables Regression can determine significant relationships between dependent and independent variables Regression can also determine strength of impact of multiple independent variables on a dependent variable Regression can also compare the effects of variables measured on different scales, such as the effect of price changes and number of promotional activities. Linear Regression Establishes a relationship between dependent variable (y) and one or more independent variables (X) using best fit straight line (also known as regression line) This is represented by an equation y = a*x + b, where a is the intercept, b is the slope of the line and e is the error term The available data is fitted to this regression line to determine the values for a, b and e. The equation can then be used to predict the value of the target variable based on any new independent variables Activity time Example 1: The regression between the experience (Years with Company) and Salary Look Look for p-value and coefficients as well Look Look for the value of R2 as it tells us how much change in Y is driven by X. Select Select your dependent variable for Y range and independent variable for X range. Click Click Data->Data Analysis->Regression Questions: What is the independent variable What is the dependent variable What if P-value is more than 0.05 What is your prediction of the salary of a person with 5 years of experience Multiple Regression This is determining the relationship between multiple independent variables and a dependent variable. The dependent variable is modelled as a function of several independent variables with corresponding coefficients, along with the constant term. Multiple regression requires two or more independent variables which is why it’s called multiple regression. It can be represented by: Y = a1x1 + a2x2 + … + anxn + b Activity time How cost variance in projects is impacted by variances in time and scope How cost variance in projects are impacted by variances in time and scope How cost variance in projects are impacted by variances in time and scope How cost variance in projects are impacted by variances in time and scope Can someone say which one has a higher impact on cost variance, scope variance or time variance Assume that P value for Time Variance was 0.25, what would you do Activity time Activity time Please copy and past this table somewhere, and remove the forecast values Please copy and past this table, and remove the forecast values Now, you can predict! That was all on Predictive Analytics!