Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 12

MBA – Industry Collaborated Programs

BusinessV Analytics (BA)

Business Analytics DISCOVER . LEARN .


Unit-III EMPOWER 1
Topics
1. Decision Tree Introduction
2. Decision Tree Making
3. Decision Tree Advantages and Disadvantages

1-2
Decision Tree For Decision Making
Decision
It is the process of choosing a course of action from among
alternatives to achieve a desired goal.
Type of decisions
Strategic Decisions
Administrative Decisions
Operating Decisions
Factors affecting decision making
Uncertainty
Complexity
High risk consequences
Alternatives
Interpersonal issues
Thus need to use a systematic and logical approach for decision
making, as it helps address critical elements, without missing
important factors. 3
Decision Tree
• A decision tree is a decision support tool that uses a tree-like model of decisions and their
possible consequences, including chance event outcomes, resource costs, and utility.
• It is one way to display an algorithm that only contains conditional control statements.
• It’s a graphical representation of possible solutions to a decision based on certain conditions
and is called decision tree because its starts with single root which then branches off into a
number of solutions like a tree.
• It is powerful and most commonly used in operations research, specifically in decision
analysis, to help identify a strategy most likely to reach a goal but are also a popular tool in
machine learning.
• Decision tree classify instances by sorting them down the tree from the root to some leaf
node, which provides the classification of instance.
• Decision tree could be used to classify as well as predict.
• E.g. Decision tree algorithms are ID3, C4.5, CART algorithms
4
Decision Tree Diagram and Symbols
Definition : A decision tree diagram is a visual model consisting of nodes and branches.
Approach : The idea is to describe several possible paths representing deliberate actions or choices, followed by events with
different chances of occurrence. By assigning probabilities and values along the way, we can evaluate each path to select an
optimal path. The evaluation is simple, consisting of alternating between calculating weighted averages or expected values at
each circle, then choosing the best action from each square. Ultimately, we obtain a value for the root node. The solution to the
decision tree consists in this pairing of root value and optimal path.
Nodes: A decision tree has only burst nodes(splitting paths) but no sink nodes (converging paths)
• Decision Node : Represented by squares
• Chance Nodes : Represented by circles and are probability nodes
• Value Nodes: Represented by triangles
Branches:
• Decision Branches are the options which come out of squares
• Event branches are the outcomes which come out of chance nodes

5
Drawing a decision tree-Example
Use case:
Manager wants to start a project A or Project B , which have an associated initial cost. Gains i.e.
revenue from project may vary depending on marketing success as very good, average or bad.
Project have their initial investment cost. Based on expected value, which is weighted average
return manager need to decide which project should she start. 4
Decision Tree making:
3 Expected Value

1.Draw the decision tree representing the options 2 Very Good


1000 * 0.3 = 300
open to the property owner Project A Average
600 * 0.5 = 300 640-500 = 140
2. Add the chance nodes and probability outcomes (500)
Bad 200 * 0.2 = 40
3. Calculate the expected values
4. Calculate the net expected value
1 Very Good
800 * 0.2 = 160
Project B Average
Key: 500 * 0.5 = 250 440-320 = 120
Decision Node (320)
Bad
100 * 0.3 = 30
Probability Node
Value Node
Do Nothing
Rejected decision 0 6
NewsExpress Problem – Part 1
• NewsExpress is a newspaper reselling company operating in the Connaught Place area of Delhi. It operates 10 stalls in the
area.NewsExpress has to buy the stocks from the newspaper companies against cash in the early morning of the day, @ Rs
3 per unit. They sell the same @ Rs 6 per unit. The unsold stocks become useless at the end of the day, and are sold in
scrap @ Rs 1 per unit at the end of the day.

• Their average daily sales were 10,000 units before the pandemic and the standard deviation was less than 100 unit

• The global Covid 19 pandemic has made their demand considerably variable and prediction now was a serious problem.
The top management was quite observant, and figured out that the demand followed a pattern and they estimated
probability of demand for various levels

Demand Probability

4000 30%

5000 40%

6000 25%

8000 5%

7
Problem 1: Estimate how much should they stock per
Solution: day
• Outcome : In case of no waste earning is 6-3 i.e. 3 Rs per unit. In case of waste the loss is 3-1 i.e. 2 Rs per unit

Thus . Outcome = Probability of Demand * (Sold units * 3 – Wasted units *2)

• Sample calculation: Scenario 5 : Stock 8000. In this case for an area only 4000 units needed and sold than 4000 are
wasted. If 6000 are stocked then 2000 are wasted. If 8000 are stocked then none wasted.

Outcome = (4000*3 – 4000*2)*0.3 =1200. Similarly For 6000 sold units it is (6000*3-2000*2)*0.25 i.e.3500

Total outcome, is some of all scenarios hence in this case 9500

• Considering outcomes for different scenario, with maximum being in scenario 2, they should stock 5000 units.
             
Scenario 2: Stock Scenario 3: Stock Scenario 4: Stock Scenario 5: Stock
Demand in Scenario 1: Stock 4000
5000 6000 7000 8000
units Probability
Outco Outco Outco Outcom Outco
Sale Waste Sale Waste Sale Waste Sale Waste Sale Waste
  me me me e me
4000 0.3 4000 0 3600 4000 1000 3000 4000 2000 2400 4000 3000 1800 4000 4000 1200
5000 0.4 4000 0 4800 5000 0 6000 5000 1000 5200 5000 2000 4400 5000 3000 3600
6000 0.25 4000 0 3000 5000 0 3750 6000 0 4500 6000 1000 4000 6000 2000 3500
8000 0.05 4000 0 600 5000 0 750 6000 0 900 7000 0 1050 8000 0 1200
   
          12000      13500      13000      11250      8
9500
NewsExpress Problem – Part 2
• Mr. Math-ur is a marketing consultant. He promised NewsExpress that he could predict the daily sales accurately for 25
days at least in a month of 30 days and has monthly fee of Rs 50,000. He also guaranteed that he would not charge any
fees if the accuracy of forecast was less than 25 days for the month. Help decide the owner of NewsExpress, on merits of
the offer of Mr. Math-ur

Solution:

• Total outcome without consultant = Average of (Highest outcome for all 30 days, Lowest outcome for all 30 days)

• Total outcome with consultant = Average (All 30 days having best prediction i.e. highest outcomes, 25 days best prediction
and 5 days lowest prediction) – the fee of the consultant i.e. 50000

• From part 1 , best outcome is 13500 and worst outcome is 9500

Case Best Case Scenario Outcome Worst Case Scenario Outcome Final outcome Decision
Without Consultant All 30 days Prediction 13500*30 = All 30 days Prediction 9500*30 = (40500+28500)
gives highest 405000 gives lowest outcome 285000 /2 = 345000
outcome
There seems to be no
With Consultant All 30 days Prediction 13500*30 = 25 days best and 5 days (13500*25 +   merits in having the
gives highest 405000 worst 9500*5 ) = consultant, as with or
outcome 385000 without consultant the
After fee(50000)   405000-50000 =   385000-50000 (355000+3350 outcome is same.
355000 = 00)/2 = 345000
335000 9
Advantages & Disadvantages
• Are simple to understand and interpret. People • They are unstable, meaning that a small
are able to understand decision tree models change in the data can lead to a large change
after a brief explanation. in the structure of the optimal decision tree.
• Have value even with little hard data. • They are often relatively inaccurate. Many
Important insights can be generated based on other predictors perform better with similar
experts describing a situation (its alternatives, data. This can be remedied by replacing a
probabilities, and costs) and their preferences single decision tree with a random forest of
for outcomes. decision trees, but a random forest is not as
• Help determine worst, best and expected easy to interpret as a single decision tree.
values for different scenarios. • For data including categorical variables with
• Use a white box model. If a given result is different number of levels, information gain in
provided by a model. decision trees is biased in favor of those
attributes with more levels.[7]
• Can be combined with other decision
techniques. • Calculations can get very complex, particularly
if many values are uncertain and/or if many
outcomes are linked.
10
Decision tree applications
• Finance Management
• Customer Relationship Management
• Fraudulent Statement Detection
• Engineering
• Energy Consumption
• Fault Diagnosis
• Healthcare Management
• Agriculture
• Planning logistics
• Strategic management
• Healthcare
11
Reference
Text Books –

1. Business Analytics; By James Evans


2. Changing Business with Data Insight, IBM Career Education
3. IBM Cognos Insight: Analyze and Share Data, IBM Career Education
Reference Books –
1. Competing on Analytics: The New Science of Winning;Jeanne G. Harris and Thomas H.
Davenport; Harvard Business Review Press; 1 st Edition (2010)
2. Fundamentals of Business Analytics; SeemaAcharya, R.N. Prasad; Publisher: Wiley India Pvt
Ltd; 1 st Edition (2011)

12

You might also like