Skip to content

Ashwani-Varshney/decision-tree

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Decision Tree

Dataset utilized- Rainfall for the Indian States

According to the dataset we have chosen for this classification problem, it contains the information about the rainfall (in cm) from Jan to Dec for different states. Here, State is the class to be predicted. We are predicting the name of the state by seeing the variation in rainfall patterns. Number of Instances: 641 Number of Attributes: 10

Library Function Usage
Pandas read_csv, DataFrame to read the dataset
NumPy Mean to calculate mean
Seaborn heatmap, distmap to visualize relationship b/w dependent & independent variable
sklearn.model_selection train_test_split to partition the dataset
Sklearn.model_selection Cross_val_score For cross validation
Sklearn.tree DecisionTreeClassifier To fit Decision Tree model
sklearn.metrics confusion_matrix, accuracy_score To obtain confusion matrix and calculate accuracy

Results-

30-fold cross validation using NB

Iteration Accuracy
1 0.5675675675675675
2 0.5151515151515151
3 0.7241379310344828
4 0.7037037037037037
5 0.6538461538461539
6 0.6538461538461539
7 0.6
8 0.56
9 0.8095238095238095
10 0.55
11 0.7647058823529411
12 0.5882352941176471
13 0.6470588235294118
14 0.6666666666666666
15 0.5333333333333333
16 0.6666666666666666
17 0.5384615384615384
18 0.5833333333333334
19 0.8888888888888888
20 0.75
21 0.5
22 0.875
23 0.6
24 0.5
25 0.5
26 0.5
27 1.0
28 1.0
29 1.0
30 1.0

Mean Accuracy 0.6813375754007939

Interpretation-

Considering the above 30-fold cross validation results, Decision Tree is not the appropriate mode l for the given dataset. Although 68.13% mean accuracy is decent but one possible reason could be the high number of labels in the class and comparatively less number of instances for each lab el which make it unsuitable.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors