Predictive Analytics Predictive Analytics

Business Analytics- Predictive analytics

1. BUSINESS A N A L YTICS UNIT V : Predictive Analytics

2. R ec a p Analytic options can be categorized at a high level into three distinct types which co-exist with, and complement each other Descriptive Analytics: Use data aggregation and data mining to provide insight into the past and answer: “What has happened?” – INSIGHT INTO PAST Predictive Analytics: Use statistical models and forecasts techniques to understand the future and answer: “What could happen?” – UNDERSTANDING THE FUTURE Prescriptive Analytics: Use optimization and simulation algorithms to advice on possible outcomes and answer: “What should we do?” – ADVISE ON POSSIBLE OUTCOMES

3. Predictive Analytics Predictive modelling is the heart and soul of business decisions. Building decision models is more of an art than a science. Creating good decision models requires: Solid understanding of business functional areas Knowledge of business practice and research Logical skills Models coevolve from simple to complex and from deterministic to stochastic It is best to start simple and enrich models as necessary.

4. Application of Predictive Analytics Online Shopping: When shopping on Amazon, predictive algorithms compare product information, purchase history and other customers' shopping activity to make prediction about what you're shopping for—and to provide personalized recommendations. Financial Services: When using a credit or debit card, bank must predict if transaction is fraudulent. It compares past history to ew purchase and denies transactions that appear to be fraudulent. Health Care: When admitted to a hospital, algorithms analyze patrient’s electronic medical record to determine if there is a risk of life-threatening infection and if at risk of illness after leaving the hospital.

5. Strategies for Predictive Decision Modeling Simple mathematics “Back-of-the-envelope” calculations can provide the basis for a more formal model Influence diagrams An influence diagram is a logical and visual representation of key model relationships

6. Implementing Models on Spreadsheets Principles of good spreadsheet design: Separate data, model calculations and model outputs clearly while designing. Do not use input data in model formulas Reference spreadsheet cells that contain data. This way, data change or experimenting with model will not easily result in errors . Building spreadsheet models (called spreadsheet engineering) is part art and part science. Spreadsheets need to be Accurate, Understandable, and User friendly . Verification is the process of ensuring that a model is accurate and free from logical errors.

7. Improve the design and format of the spreadsheet itself. Sketch a logical design of the spreadsheet. Break complex formulas into smaller pieces Design the spreadsheet in a form that the end user can easily interpret and understand Improve the process used to develop a spreadsheet. Work on one part at a time Check formula results with simple numbers Inspect your results carefully and use appropriate tools available in Excel. Use Excel auditing tools Improving Spreadsheet Quality

8. Spreadsheet Applications in Business Analytics A wide variety of practical problems in business analytics can be modeled using spreadsheets. A useful spreadsheet model need not be complex; often, simple models can provide managers with the information they need to make good decisions.

9. Logic Driven Predictive Model: Multiple Time Period Model Many practical models incorporate multiple time periods that are logically linked together. Taking a systematic approach to putting the pieces together logically can often make a seemingly difficult problem much easier.

10. Single Period and Multi Period Models Single-period inventory model One-time purchasing decision (e.g., vendor selling T-shirts at a football game) Seeks to balance the costs of inventory overstock and under stock Multi-period inventory models Fixed-order quantity models Event triggered (e.g., running out of stock) Fixed-time period models Time triggered (e.g., monthly sales call by sales representative)

11. Logic Driven Predictive Model: Single Period Purchase Decision Overbooking of airline flights Ordering of clothing and other fashion items by a Boutique, Big Billion Day Sale One-time order for events – e.g., t-shirts for a concert

12. Logic Driven Predictive Model: Single Period Purchase Decision Consider the problem of deciding how many newspapers to put in a hotel lobby Too few papers and some customers will not be able to purchase a paper, and profits associated with these potential sales are lost. Too many papers and the price paid for papers that were not sold during the day will be wasted, lowering profit.

13. Logic Driven Predictive Model: Overbooking Decision United Continental sparked anger in social media in April 2017 after a video showed a passenger being forcibly removed from an overbooked flight. Overbooking is common practice for the majority of airlines and is perfectly legal. Airlines sell too many tickets with assumption that there will “no shows” — either because people miss flights, previous flight is delayed or they change travel plans. The statistical chance of all passengers with a valid ticket checking in on time is less than 1 in 10,000 at best. Compensation depends on the distance and can range from $250 for short-haul flights to $600 for long-haul trips to a maximum of around $1350.

14. Sources of data Subjective judgment Existing databases Analysis of historical data Surveys, experiments, or other methods of data collection Data and Models

15. Data Driven Predictive Models: Retail Markdown Retailers manage revenue by reducing product price based on season or timing Idea is, as season goes on, value to customer falls – esp. seasonal, perishables Markdowns are discount made on merchandise in retail store from original sale price Compared to a sale or promotional event, a markdown (in its purest form) is when you change list price to a lowered price permanently Whatever be the product or service, it is important to understand the trend and seasonal aspects of product sales. Use moving averages to eliminate seasonality to see the trend in sales Use solver to develop additive or multiplicative model to estimate trends & seasonality There are three variables in the model Base, Trend and Seasonality Index.

16. Mahindra is all set to launch its new flagship model, the XUV700, on 9th October. Ahead of XUV700’s launch remaining stocks of Mahindra Rexton receives massive discounts of up to Rs 9.5 lakh! You can also get a TUV300 if not interested in cash discount.

17. What if analysis What-If Analysis is the process of changing the values in cells to see how those changes will affect the outcome of formulas on the worksheet. Three kinds of What-If Analysis tools come with Excel: Scenarios, Goal Seek, and Data Tables.

18. What if analysis Scenarios and Data tables take sets of input values and determine possible results. Data Table works with only 1 or 2 variables, but can accept different values for variables. Scenario can have multiple variables, but it can only accommodate up to 32 values. Goal Seek works differently from Scenarios and Data Tables in that it takes a result and determines possible input values that produce that result. In addition to these three tools, add-ins that help perform What-If Analysis, such as the Solver add-in . Solver add-in is similar to Goal Seek, but it can accommodate more variables. Help create forecasts by using fill handle and various commands that are built into Excel.

19. What-if analysis One-Way Data Table Build the one-way table, reference the output in the adjacent column Select Table range and then select data table, in column input cell reference demand value Two-way Data Table Build a two way data table, with unit cost and unit price in row and column Select Table range, then select data table, in colum input refer Price and row input refer cost

20. END OF UNIT V