You are on page 1of 3

WINE DATA SET MODELLING- BABI Group 7 Project

Submitted By:

GROUP - 7

X014/14
X023/14
X017/14
X020/14

HARISH DEEPAK
PRAMOD KUMAR
MANOJ KUMAR
NEYAZ AHMAD

1. Deliverables: With the objective of developing an efficient classification model from


the project wine Dataset for modeling the group worked using Neural network in on
XLMiner and SPSS Modeler platform. The second task is to classify the data given in
Wine Dataset for evaluation and to assess the accuracy of the classification model once
the actual class of the data are known.
2. Methodology:
(i)
A study was conducted using Wine Dataset for modelling which contains 13
variables and 1 class variable. It was also not obvious that all these 13 variables are
truly independent or uncorrelated to each other. So to obtain result we used Excel
Miner and used Classification tree/Single tree and inputs of all 14 variables were
provided for entire 120 datasets and the data set training data, validation data and test
(ii)
(iii)

data was provided.


Model Development: Excel Miner and SPSS Model was used.
Post Model Development: We were provided another data sheet without defined
class. Both the sheets were opened simultaneously and in second sheet we used the
tool score and from the drop down changed the file to model file and compared both
the variables by name.

XLMiner : Neural Network


Classification
Parameters/Options
Input variables normalized
Network Architecture
Seed: Initial Weights
# Hidden Layers
# Nodes in Hidden Layer 1
# of Epochs
Step size for gradient descent

Yes
Manual
12345
1
25
30
0.1

Weight change momentum


Error tolerance
Weight decay
Cost function
Hidden layer activation
function
Output layer activation
function

0.6
0.01
0
Sum of
squares
Standard
Standard

Output Options Chosen


Summary report of scoring on training data
Detailed report of scoring on training data
Lift charts on training data
Summary report of scoring on validation data
Detailed report of scoring on validation data
Lift charts on validation data

Prior Class Probabilities


Class
Premium
Regular

Prob.

0.402778
0.597222

Training Data Scoring Summary Report


Confusion Matrix
Predicted Class
Actual
Class
Regular
Premium

Regular

43
0

0
29

Error Report
Class
Regular
Premium
Overall

# Cases

43
29
72

# Errors

% Error

0
0
0

0
0
0

Performance
Regular
1
1
1
1

Success Class
Precision
Recall (Sensitivity)
Specificity
F1-Score

Validation Data Scoring Summary Report


Confusion Matrix
Predicted Class
Actual
Class
Regular
Premium

Regular

Premium

25
0

2
21

Error Report
Class
Regular
Premium
Overall

# Cases

27
21
48

# Errors

% Error

2
0
2

Performance
Success Class
Precision
Recall (Sensitivity)
Specificity
F1-Score

Regular
1
0.925926
1
0.961538

7.407407
0
4.166667

You might also like