MIS 420 Predictive Analytics Making Statistical Conclusions analytics

MIS 420: Predictive Analytics & Making Statistical Conclusions

[analytics: recap]

[descriptive analytics] “Using descriptive analytics is like driving your car but watching traffic through the rear-view mirror, not seeing what’s ahead and thereby in danger of crashing. ”

[predictive analytics] “Using predictive analytics is like driving your car and watching traffic through the front windshield, anticipating traffic, making course corrections to avoid traffic jams and getting there faster and safer. ”

[predictive analytics: definitions] “exploit patterns found in historical and transactional data to identify risks and opportunities … capture relationships among many factors to allow assessment of risk or potential associated with a particular set of conditions, guiding decision making for candidate transactions. ” “predicting future events and assessing the attractiveness of various courses of action. ” HELPS TO ANSWER: WHAT SHOULD WE DO NEXT? ? ?

[descriptive vs. predictive analytics] • Descriptive Analytics: the understanding of existing (retrospective) data with the goal of understanding trends via comparison • This is the first step towards deriving predictive analytics • Predictive Analytics: more sophisticated analytics that “forward thinking” in nature • gaining insights by enhancing understanding, interpretation and judgment for the purpose of good decisionmaking • Remember, not what WILL happen, but what is probable to happen, based on past data.

[where are predictive analytics used? ] • Marketing • Financial services • Insurance • Telecommunications • Retail • Travel • Healthcare • Pharmaceuticals • …. And more!

[communication skills and predictive analytics] • Not enough to have state of the art in BI technologies, without having a common understanding and a common language between the business people and BI professionals, otherwise BI efforts will fall short of desired results • Business professionals need to appreciate the role of technology as an enabler and they need to lead and determine where & how IT/BI infrastructure should be deployed to improve decision making • IT & BI professionals need to understand the language of strategy, business models and performance while solving business NOT technology problems

[benefits of predictive analytics] • Informs decision making • Rapid identification of new market opportunities • Increased confidence of pursuing new and existing business strategies • Enhanced ability to direct and focus decisions • Draws reliable conclusions about current conditions and future events • Timely feedback to executives on their strategic initiatives – without feedback course corrections may be too late • Provide leading indicators and insight to assist in planning for answering the big question: What should we do next? – next quarter, next year etc.

[overall: why do organizations need predictive analytics? ] • Improving business processes • Doing more with less budget (working smarter not harder!) • Allocating resources appropriately • Understanding correlations and sensitivities with customer segments • To ensure long term financial resources are available to run the business • Developing Competitive Advantages and Differentiation

[EXAMPLES IN TABLEAU]

[Predictive Analytics in Tableau: Understanding Trend Lines] • Trend Lines: indicating the general course or tendency of something • Based on linear regression • Example: Want to look @ relationship between Sales and Profit. • For Central Region (blue trend line), as Sales increase, so does Profit. So, based on historical data, as sales increased, profit tends to increase. • How do we know if this trend line is meaningful, or legit? • Look @ Trend Model

[Predictive Analytics in Tableau: Model Formula] • Example: Right-click on the Central trend line > select Describe Trend Model • Model Formula: Profit = Coefficient for Region*Sales + yintercept • Y = mx + b • Y-intercept: is a point in the equation where the x-value is zero. • Formula for Central: • Profit =. 198449*Sales + 10. 7668 • Basically, Tableau plots the line for every piece of data using that formula

[Predictive Analytics in Tableau: Coefficients, YIntercept, Interpretation] • Formula for Central: Profit =. 198449*Sales + 10. 7668 • Coefficient for Central: . 198499 • What does Coefficient mean? ? • For every unit increase of X, Y goes up by the • coefficient. • In our case: • In the central region, for every dollar our Sales increases, our Profit increases by $0. 1984 cents. • What is the formula for East? • • Sales coefficient: . 154 Y-Intercept: . 385 Profit =. 154*Sales +. 385 In the Eastern region, for every dollar our Sales increases, our Profit increases by $0. 154 cents.

[Predictive Analytics in Tableau: P-Value] • Formula for Central: • Profit =. 198449*Sales + 10. 7668 • • • • Meaning: In the central region, for every dollar our Sales increases, our Profit increases by $0. 1984 cents. Is this significant? Need to look @ p-value P-value: way to measure significance How likely result would occur by chance. A calculated P value of. 001, we can say that the observed outcome would be expected to occur by chance only 1 in 1, 000 times Criteria: Standard cutoff is less than. 05 P > 0. 10: not significant 0. 05 < P < 0. 10: weak evidence 0. 01 < P < 0. 05: moderate evidence 0. 001 < P < 0. 01: strong evidence P < 0. 001: very strong evidence So is our statement meaningful for the Central region? P-value < 0. 0001 • So we can conclude that: • In the central region, for every dollar our sales increases, our profit increases by $0. 1984 cents. There is very strong evidence for this statement, given that p < 0. 00001.

[Predictive Analytics in Tableau: Practice] • Now using the same table, answer the following questions for the West Region: • What is the coefficient? • What is the y-intercept? • What is the formula for the West region? • What does the coefficient mean in terms of profit and sales? • What is the p-value for the West region? • Can we make significant conclusions for the relationship between sales and profit in the West? Why or why not? (use p-value to support your answers)

[Predictive Analytics in Tableau: Forecasting] • Forecasting: estimations of future values of a measure based on historical values • Requirements: a measure, and a continuous date field. • A continuous date is one that has an infinite number of possible values. For example, the Order Date field in Superstore Sales has possible values from year to individual day • The measure must be a basic aggregation—forecasts can’t be calculated for table calculations • At least one complete cycle of the data that you want to forecast—for example, if you want to forecast data for one year, you need at least one year—preferably two or more—of data

[Predictive Analytics in Tableau: Forecasting Example] • Example: let’s forecast sales over time • Drag Sales into Rows • Drag Order Date into Columns • Change Order Date from Year to MONTH • We now have a nice line graph that shows actual data. Let’s forecast what the data will look like over the next year •

[Predictive Analytics in Tableau: Forecasting Example] FORECAST • Example: Sales over time • Right click on graph > Show Forecast • Let’s play with some of the forecast options.

[Predictive Analytics in Tableau: Forecast Options] • Example: Sales over time • Right click on graph > forecast options • Forecast Options are basically the assumptions of your forecast • Forecast Length: • • Notice Auto is set to Next 13 months Can Change it to whatever you want to forecast • Source Data: • The Aggregate by – you can change the aggregate as well

[Predictive Analytics in Tableau: Describe Forecast] • Example: Sales over time • Right click on graph > Describe Forecast • • Sum of Sales: value being forecast (this will change if your data changes) Initial: prediction interval for first forecasting period Our first is Dec 2013 (the interval is +/- $123, 597) Change from Initial: difference between first and last forecast estimates Seasonal Effect: These fields are displayed for models identified as having seasonality--that is, a repeating pattern of variation over time. Contribution: The extent to which Trend and Seasonality contribute to the forecast. These values are always expressed as percentages and add up to 100%. Ours is only trend because no seasonality was indicated Quality: how well the forecast fits the actual data. Possible values are GOOD, OK, and POOR. A naïve forecast is defined as a forecast that estimates that the value of the next period will be identical to the value of the current period. Quality is expressed relative to a naïve forecast, such that OK means the forecast is likely to have less error than a naïve forecast, GOOD means that the forecast has less than half as much error, and POOR means that the forecast has more error

[Predictive Analytics in Tableau: Things to Remember in Forecasting] • The forecast will only ever be as good as your data, and the more you have of it, the better the forecast should be. • The forecast can’t account for external influences on your data, though, but you can factor those into your historical data to build more complex forecasting models. • For more information on forecasting, check out • • • http: //beta. tableausoftware. com/qsg/forecasting. pdf (username: v 8 beta password: kraken http: //beta. tableausoftware. com/videos/Forecasting/Forecastin g_Tableau_8_Beta. html (use the username and password above) http: //onlinehelp. tableausoftware. com/v 8. 0/pro/online/enus/help. html#forecasting. html