Activities/Assessment:

1. What are the steps being undertaken in the predictive analytics phase of the

business analytics process?

Businesses must be cautious when choosing data partners and guaranteeing that

their data is accurate and not segregated or otherwise limited. It is simple to get

started with external data that has been validated and can reveal other factors that

contribute to patterns that change quickly thanks to a variety of data providers.

Although one expert may be in charge of each of these procedures, each step of the

process should be seen as a collaborative one.

Step in predictive analysis stage in business analytics process.

Project Initiation

- You must first begin the project in order to use the predictive analytics

process. Finding the right employee to sponsor the effort, gain their

approval, and get things rolling will be the first step.

Project Requirements

- In order to solve problems like reducing fraud, maintaining inventories,

improving customer suggestions, or increasing the value of a loan

portfolio, predictive analytics typically begins with a business expert. The

next step is to decide what the project will comprise and how it will be

completed when the project team has been constituted. Business needs

should be the primary consideration when creating the requirements for a

modeling project, followed by technical requirements. Business

requirements outline the goals and strategies for the organization, as well

as any challenges that must be overcome.

- Analyze the data to see if it fits the criteria of the aim. To improve

predictions and achieve other business objectives, a data analyst should

ascertain what data sets are available and how to use them. One

technology that can be utilized to address specific business problems is

predictive analytics. It's a fantastic tool, but it's not always the best or the

best option. It's time to decide whether predictive analytics is the best

choice once you have a clear understanding of the project's requirements.

Before the data is fed into the predictive analytics model, any problems

may be detected if it is known how and why the data is gathered.

Model Building and Business Evaluation

- Once the technical specifications have been established, the data

scientist can start building the model. There are many tasks involved in

building models. Choose the predictive models that will yield the most

useful results, and then work backward to determine the best method for

arranging the raw data into the correct set of attributes for the model. To

discover the ideal balance of performance, accuracy, and other

requirements like explain ability, experiment with a variety of qualities,

approaches, and practices. The first is the Extract, Load, and Transform

(ELT) process, which entails gathering data from pertinent data sources,

putting it into the analytical environment, and formatting the data

appropriately.

Implementation
 - Right from the outset of the project, the implementation strategy for

predictive models should be taken into account as part of the

requirements stage. The newly created predictive model needs to be put

into production in order to produce results. A unique predictive analytics

finding is only helpful if it materially alters the ongoing process. Numerous

models have been delayed or never implemented as a result of a lack of

planning for their deployment.

Monitoring and Redevelopment

- The effectiveness of predictive models decreases over time. When a

model has served its purpose, a newer, more modern model must be

used in its place. Model degradation is probably not a problem if a

predictive model is used tactically once, but you should still evaluate how

well the model worked and apply what you discover to future projects.

Model deterioration should be a top priority if you plan to utilize the model

frequently to guide important corporate decision-making processes.

Model decay is a problem that is occasionally disregarded but is one of

the most crucial aspects of model application.

How Long Should a Predictive Analytics Project Take?

- Depending on the scope and context of what you're trying to do, the

industry you're in, and the cultural approach to project management, risk,

and regulatory monitoring, the time and effort required to develop, build,

and implement a new predictive analytics application will vary. If the data

science was successfully completed, the predictive analytics model

should satisfy the performance, accuracy, and other requirements when it

time. Model performance may be impacted by alterations in customer

sentiment, the business environment, and other variables.

2. Who are the business organizations’ project sponsors? Explain their role in the

predictive analytics process

A project sponsor is a person or organization who funds the project and offers

support to make sure the portfolio, program, or project is successful. Every project

needs at least one project sponsor, so in order to get things rolling, you'll need to

choose the best candidate inside the organization to support the initiative, receive their

approval, and get things going. The project sponsor or senior executives should be

someone who can approve the funds and resources required to complete the project

as well as someone who can defend the project if it is contested by other senior

managers for a new predictive analytics application, one that is strategically

significant, or one that is contentious. If the model creation is more tactical or low-key

in nature, or if you are just replacing an outdated model with a more current version,

the project sponsor will most frequently be the head of the department implementing

the model.

Project Managers

This person ensures that the project's key milestones and objectives are met

on time and at the required standard. This is the main point of contact, and its tasks

are as follows:

Acquiring and administering resources

Ensuring that the company's procedures are observed

Keeping track of how well the plan is working and track the difficulties as the

project progresses

Decision Makers

These are the people in positions of power who make crucial business decisions:

Prior to deployment they need to obtain budget approval and signing off on

the model

The project sponsor frequently gives them authority to act on their behalf.

They have the ability to produce real-time insights.

They can also forecast how to improve their performance.

Data Scientists

The subject matter expertise for analytical techniques, data modeling, and

applying accurate analytical approaches for a specific business challenge is

provided by a data scientist. They will take part in the following tasks:

Data collection and analysis, as well as model construction

Requirements definition, testing, implementation, and documentation

production

Ensuring that the overall analytic goals are met

They use analytical approaches and work with the data that is available for

the project at hand

Stakeholders
 Stakeholders are people who are affected by the model. A stakeholder's

responsibilities include the following:

Full-time dedication to project management and implementation in order to

translate the chosen corporate strategy into action and activities.

Determines and controls the business case's strategic elements

The total project pipeline is prioritized

Establishes and oversees project delivery techniques and standards, as well

as keeping track of project progress

3. Identify and describe the project’s business and technical requirements; and

explain the importance of these requirements in the accomplishment of the

predictive analytics process.

Technical requirements are generally focused with analytical issues, whereas

business requirements outline the goals and strategies of the company as well as any

challenges that must be overcome.

Predictive Analytics Process's Technical Requirements:

Modelling Objectives

- The modeling aim is the behavior that the model forecasts will occur. In

statistics and data mining, the term "objective function" is sometimes used

to refer to the modeling objective. This is important because the modeling

objectives define the business issue in terms of individual-level events

and non-events.

Outcome Period (Forecast Horizon)

because doing so enables you to understand potential outcomes.

Data Sources and Predictor Variables

- This section explains the source of the model's data, which is important

because data scientists may use many data pieces in predictive models.

It also specifies which predictor variables should be taken into account

while developing the model.

Sample Definition

- The process of deciding which subset of the available data should be

used to create the model is known as sample definition. By doing so, the

precise data that will be used to create, train, and adjust the statistical

model as necessary to deliver accurate results will be identified.

Model Type and Method of Construction

- Any model type is acceptable, and any modeling approach is acceptable,

however this criterion requires a specific sort of model that was

developed using a specific technique.

Business requirements outline the objectives of the organization, how they will be

achieved, and any challenges that must be overcome.

Business Requirements in the Predictive Analytics Process:

Business Objectives

- These are the results that the project will deliver to help the company

accomplish its goals. This will make it easier to anticipate increased

the project that is directly tied to how the business operates.

Scope

- The project's direction will be determined by this necessity, which will also

dictate what will and won't be covered by the project.

Legal Requirements

- This is an important necessity since businesses must abide by consumer

legislation in order to prevent rule violations, protect the project from

fines, and avoid legal action.

Implementation Path

- The model's integration with the business processes of the company is

discussed in this section so that it can be used on a regular basis. By

doing so, it will be possible to determine whether certain project needs

will have an effect on call centers, IT, or other divisions of the business.

Project Planning

- The business will inquire about the project's price and duration. If these

specifics aren't worked out beforehand, very few organizations will agree

to take part in a project.

Risks and Issues

- Risks are potential outcomes that endanger the project's success but

aren't guaranteed to occur. Issues are problems that have already

happened and are impeding development. Risks should be reduced and

problems should be fixed.

Documentation and Reporting

information to stakeholders. It could be necessary to provide extensive

paperwork to comply with internal/external audit procedures.

The focus of business requirements is on organizational goals, governance, and

control. Technical needs address data and analytical concerns. The business and

technology requirements must, however, be compatible, and the technical

requirements must be able to be linked to the business requirements.

4. What do you think are the reasons for not pushing through with any predictive

analytics project?

It's very easy to get predictive analytics poorly. Because predictive analytics is only one

technique that can be used to solve certain sorts of business issues. It can be an

excellent tool, but it is not necessarily the best or the most appropriate tool.

Here are some of the reasons why certain predictive analytics projects failed:

Lack of Stakeholder Buy-In

- If the project sponsor and key stakeholders are unfamiliar with and

specialized in predictive analytics, the project is likely to fail.

No purpose

- If the model has no practical business use, it is not necessary to have it.

Making a model is useless if no one will use it.

Lack of Data

- Without data, you cannot do a data science project. Along with having at

least a few hundred samples of the behavior you're attempting to predict,

time.

Deterministic Outcomes

- A predictive model is not necessary if you can confidently predict what

someone will do or act like.

Over-expectation

- It's unlikely that the model will live up to its claims. Predictive analytics

can boost efficiency by 20–30% and/or the company's bottom line. There

is a danger the program will fail if the aim is to produce more benefits

than this.

Accurate but not precise

- It is possible to construct a workable model, but it won't be accurate

enough for the task. The model must be able to pinpoint those with a 95%

chance of occurring in order for it to be useful.

No implementation route

- Because there isn't a means to achieve it, the model cannot be put into

practice. Costs of execution, prioritization difficulties, or legal obstacles

are frequently to blame for this.

5. Enumerate and explain the meetings required for model building and business

evaluation.

It is more crucial for the business to have a clear and thorough grasp of what the

data scientist has done and to guarantee that the right model has been constructed to

suit their objectives the greater the model's potential effect and strategic importance.
 Several meetings between the data scientist and stakeholders are anticipated during

the phases of model development and evaluation for a model that will be used to

choose how patients are treated or to make daily insurance decisions worth millions of

dollars.

Meetings between the data scientist and stakeholders during the model-building and

evaluation phases.

1. Data Meeting – This is where we’ll talk about some preliminary findings on which

data can predict behavior. The data scientist should bring up data items that

indicate a very strong uncommon, or unexpected association with behavior

during this meeting.

2. Preliminary Results Meeting – The data scientists discuss the models they’ve

created at this meeting. In many cases, it’s useful to talk about which predictor

variables are included in the model and how much they contributed to the final

score.

3. Evaluation/Usage Meeting – This discussion will go over the various cut-offs that

could be used and how they would affect the business.

6. Explain the importance of monitoring and redevelopment in the predictive

analytics process.

Predictive analytics data scientists have quick access to a wide variety of statistical,

data-mining, and machine-learning methods created for use in predictive analysis

models. However, it is evident that the performance of prediction models deteriorates

over time. When a model has served its purpose, a newer, more modern model must

be used in its place. Based on the results, you can enhance an existing model or give

the system a new capability. The customized model was developed to handle a
 particular business challenge or collection of difficulties. Keep an eye on the predictive

analytics project so that you can quickly update it with new data that enables the

development of more predictive models whenever new or better data becomes

available. While you're monitoring, there can be a paradigm change or a data shortage.

Support teams are now better equipped with predictive monitoring technologies and

resources to identify root causes even when an event happens suddenly or quickly.

7. What do you think is the ideal timeline in accomplishing the predictive analytics

process?

A significant portion of this time will be spent on documentation and validation in

order to satisfy the model's user, demonstrate that the models are stable and robust,

and demonstrate that you understand the model's impact on the project. The time and

effort necessary to plan, construct, and implement a new predictive analytics

application will vary depending on the scale and context of what you're attempting to

do, the industry you're in, and the cultural approach to project management, risk, and

regulatory oversight. Here are some of the scenarios in predictive analytics project and

the timeframe required to accomplish the specific scenarios:

Scenario Time Required

If the data is in perfect condition and Few Minutes

Gathering the data, preparing data, and 2-4 Months

After they've developed the model, you'll need to put it into action. When comparing

timeframes offered by internal and external developers, be sure that both quotes

cover the same set of deliverables