Activities/Assessment:

1. In your understanding, what are the core performance areas of a well-

structured and properly executed Predictive Analytics program?

Businesses may use data to better understand their customers, optimize their
advertising efforts, tailor their content, and boost their profits. The benefits of data
are numerous, but you can't take use of them without the right data analytics tools
and methods. While raw data has a lot of potential, data analytics is required to
harness that potential. Predictive analytics is a broadening of present tools and
approaches such as the performance or balanced scorecard, forecasting, and target
setting, to name a few. It is vital to have knowledge and awareness of the activities
within the organization in order to build metrics across all functions and how they
interrelate before beginning to forecast trends and behaviors that have a financial
impact. Internal business users' expectations for management data are rising, both in
terms of the data to be analyzed and the amount of analysis required. From huge
concerns to small ones, all leaders, managers, and operational employees expect
better information to help them make better decisions.

2. What do you think is the role of a Predictive Analytics business partner?

An individual or a specialized function within a company that will provide

analytical support to decision makers and others is referred to as a Predictive

Analytics business partner. These specialists are increasingly expected to be able to

provide organizational leaders and managers with information and analysis on the

organization's condition and trajectory. They take part in strategic and operational

decision-making and are willing to ask pointed questions when necessary to ensure

that the organization is operated in the long-term best interests of stakeholders.

These experts must contribute to earn their reputation as valuable business partners,

and their contributions must be founded on true numbers and facts.

is possible to work on any business analytics project without having to use

predictive models? Why or why not?

Data, statistical algorithms, and machine learning techniques are used in

predictive analytics to determine the likelihood of future outcomes based on historical

data. The purpose is to provide the best judgment of what will happen in the future,

rather than only knowing what has happened. When we discuss efficiency, we're

usually talking about automating a manual decision-making process. People are

occasionally reassigned to more productive roles as a result of this, but efficiency

more often than not results in job losses and/or a devaluation of people's skills. This

is significant because it means that if you're implementing predictive models for the

first time, or in a new area where they've never been used before, you'll likely face

resistance and will need a strategy to deal with it. In terms of the second point,

multiple studies have shown that predictive analytics algorithms provide better

forecasts than humans, and in many circumstances, better predictions equal more

money.

4. What is the role of decision-making in constructing predictive models?

The purpose of predictive analytics is to figure out the relationships between

predictor and behavioral (outcome) data. Predictive analytics is impossible without at

least one of these categories of data. You'll need representative samples of each

type of activity for behavioral data in order to examine differences between

behaviors. For the sake of our case study, this means that information about people
 who did not buy wine is just as valuable as information about people who did. The

data is then examined using the predictive analytics method to evaluate how the

predictor data may be used to distinguish between various actions. In an increasingly

customer-oriented world, businesses have accumulated a lot of consumer

knowledge and data. In order to stay competitive, businesses must leverage

customer data to shape their products, solutions, and purchasing experiences.

Companies that strategically leverage predictive consumer behavior insights

outperform their peers by 85 percent in sales growth margins and by more than 25%

in gross margins, according to Mckinsey study. As a result, understanding the

strategic value of consumer data is crucial for executives.

5. What are the different types of predictive models and when can these models

be used?

 Linear Models

The most often used predictive analysis technique is linear

regression. It forecasts the future of a dependent variable (target) using linear

relationships between one or more independent variables (predictors). The

prediction is based on the premise that the target and predictors have a

causal or dependent connection. The dependent variable and the

independent variables are assumed to have a linear relationship in linear

prediction models. As a result, these models have a lot of bias and a lot of

variability. Because of the assumption of nonlinearity, these models have a

significant bias.

 Decision Trees
 The decision tree is a common predictive modeling algorithm that may

be used to visually and explicitly express decisions. It's a graphical

representation that employs branching methodology to illustrate all possible

outcomes in response to given variables. The internal node of a decision tree

represents a test on the attribute, the branch reflects the outcome, and the

leaf represents the conclusion made after computing the attribute. Decision

Tree is a tool that aids in decision-making and communication in specific

situations. It aids data scientists in grasping the concept of how different

decisions can result in different operational aspects of a scenario. It assists in

making the best choice. When instances are represented by attribute values

and training data contains error, the technique is well suited. It can also be

used when the target function's output value is discrete.

 Neural Networks

A neural network is a sophisticated computer data model that can

represent and capture complicated input/output relationships. Artificial neural

networks are forecasting techniques based on simple brain mathematical

models. Complex nonlinear relationships between the response variable and

its predictors are possible with them.

 Support Vector Machines

For two-group classification issues, a support vector machine (SVM)

is a supervised machine learning model that employs classification

techniques. SVM models are capable of categorizing new text after being

given sets of labeled training data for each category. They have two key

advantages over newer algorithms such as neural networks: they are faster
 and perform better with less samples (in the thousands). This makes the

technique ideal for text classification tasks in which just a few thousand

tagged examples are available.

 Cluster Models

A data cluster is a machine learning algorithm that groups data into

sets with similar properties in order to generate data models. By forecasting

future behavior or results of a given cluster, data clusters are one modeling

path for predictive analytics. Cluster models are what they sound

like. Clustering models let you divide data into a set of groups. This can aid in

the discovery of natural groups within your data. Clustering models are

concerned with discovering groups of related records and categorizing them

appropriately.

 Expert Systems

Expert systems represent experts' problem-solving thinking through

rules. The criteria are based on understanding of the issue domain and

techniques. Textbooks, research articles, interviews, surveys, and protocol

analysis are all good places to start when learning for an expert system. If the

topic to be modeled is complex, or if specialists are unaware of their own

procedures, protocol analysis is extremely effective. Expert systems should

be simple to use, include the most up-to-date information, and explain why

they make suggestions. Expert systems are particularly promising in

forecasting for replacing unaided judgment in circumstances needing multiple

forecasts, modeling complex problems with insufficient data on the

dependent variable, and handling semi-structured problems. In two

judgmental bootstrapping. Expert systems and econometric models were

compared with minimal evidence; expert systems were better in one study

and tied in two others.