Professional Documents
Culture Documents
AND BUSINESS
APPLICATIONS
Predictive analytics or Predictive
modelling
MONITORING PREDICTIVE
DATA ANALYSIS THE MODELS ANALYTICS
Predictive Analytics Process
TYPES OF PREDICTIVE
ANALYTICS
Machine
Manpower
(hoop & Figure 6.7
(shooter)
backboard)
Single period Purchase problem is
another example for Logic driven model
Decision to be taken on how much quantity of newspaper or Valatines cards to be purchased by a
retailer. It depends on the experience about Demand and discount rate. Some quantity can be
sold at normal price R and unsold quantity has to be sold at a Discount price.
If Purchase Price is C
Selling normal Price is R
Selling Discount price is S
Quantity purchased is Q
Expected quantity demanded is D
Then Profit = Sales value – Purchase cost
Sales value = R x Quantity sold + S x Surplus quantity
Purchase cost = Q x C
Profits = Sales value – purchase cost.
The seller uses his experience to fix a Demand and Discount rate.
This model can predict profits for any given value of D and Q.
This model can be implemented in Excel Spreadsheet.
DATA DRIVEN PREDICTIVE MODEL
These models can predict future outcomes quantitatively using past data.
Eg. Sales forecasting models where past data on Sales and Advertisement are
available. Regression model, Time series model.
Eg. Retail Pricing Mark down model
In Retail shops, new stocks are sold at a certain price. After some time the
demand comes down or becomes out of fashion. So the balance stocks are to
be sold at a mark down price or Discounted Price. This Mark down gradually
increases weekly or daily till all stocks are sold.
The Retail Pricing Mark down model can predict the timing when the mark
down to be announced and what level of mark down to be offered each day or
week so that total stocks can be cleared over a particular time period. The
model uses past data of quantities sold at various price levels.
Eg. A store has 1000 units of Umbrellas in stock. They want to clear the stock
in a time period of 1 year. Using past year sales data and price, the Model will
advise when to start the discount sale and how much discount to be offered
every week, till the end. It can also estimate the Profits earned.
MODELLING RELATIONSHIPS OR FORMULA
IN PREDICTIVE ANALYTICS
LINEAR FUNCTION Y = A + BX.
Eg. Linear Regression model where Sales Y can be predicted as a
Constant A + B x Advertisement budget. From past data on
advertisement and sales, we can predict the future sale in any
year, if the advertisement budget is given.
Polynomial function- function with square or cube variables.
Eg. Multiple regression model Y = a + bX + c z2 + d k3
Logarithmic Function Y = log X
Power function Y = axb
MODELS INVOLVING UNCERTAINTY
WHAT IF ANALYSIS
SENSITIVITY ANALYSIS
GOAL SEEKING ANALYSIS
OPTIMISATION ANALYSIS
WHAT IF ANALYSIS
AN ANALYSIS THAT HELPS TO FIND OUT , HOW MUCH THE OUTCOME WILL
CHANGE IF ONE INPUT OF THE MODEL CHANGE BY CERTAIL %.
Eg. A model that estimates how much the inventory cost change if the
carrying cost change by 2 %
How much the total profit change if the Bank loan interest change by 2 %.
How much the Sales will change if the advertisement budget increased by
10%. ( Advertisement elasticity)
Tools in Excel for What – If analysis
Scenarios. Excel saves each % or change in a variable as a scenario. We can
ask Excel to compute Sales for each scenario of Advertisement increase
10%, 15%, 20% etc.
Sensitivity analysis
It is similar to What if analysis.
It studies the change in output of a Model, for a very small change in input
variable.
Automatic sensitivity analysis allows only one variable change at a time. It
provides quick results.
Trial and error sensitivity analysis. It permits changes in more than one
parameters. By making several changes in variable values and estimating
the output, by trial and error, we get better and better solutions.
Goal seeking analysis
It determines the value of inputs required to achieve a certain output or
achieve the goal
Eg. To achieve 20% growth in Profits, how much advertisement budget is
required?
To reduce Patient waiting time in a Hospital to 10 min, how many Doctors
should be available?
Goal seeking analysis can be considered the reverse direction of What if
analysis.
In What if analysis, an input is varied and its effect on the output is
studied. In Goal seeking analysis, the change in output required is decided
and the corresponding input required is studied.
Both these can be done in Excel
Optimisation analysis