thesis formatting and finalizing writing
Budget: $30 – $250 USD
Summary of my project :
This research is about prediction for customer satisfaction using feature selection compare it with all attributes accuracy
We have 2 datasets:
First kano dataset ( kano dataset aug ); we use it for kano feature selection based on categorizing feature to 5 categories as following
1- must be
2- one dimension ( performance)
3-excitement
4- indifferent
5- reverse
Feature located under 2 and 3 categories we consider them as important and we will call them kano features so we compare them with other feature selected by feature selection method to see how is the prediction with different scenario ( take the common between them r union or kano +1 or kano +2 or kano +3 or cluster to male and female or 5 cluster or 8 clusters using kano dataset then use the same records in satisfaction dataset to cluster the data
The second dataset is satisfaction dataset the target column is satisfaction level from 1 to 5 . the name of the dataset is( satisfaction dataset 2022 )
Feature can be selected using kano and compare with other feature selection methods ( anova , lasso , natural information and chi square )
Prediction use different methods ( xgboost regression, random forest , adaboost , Decision tree and multi linear regression
Experiments include prediction using -
1- all attributes
2- kano ( one dimension and excitement )features the ry are 9 features
3 - features selected
4- common between kano and feature selection
5- union between kano and feature election
6 union kano + 1 from feature selection method
7- union kano + 2 from feature selection method
8- union kano + 3 from feature selection method
9 - cluster to make and female and predict using all and common
10 - cluster to 8 clusters and compare the feature match between. All cluster using different selection method
11- - cluster to 5 clusters and compare the feature match between. All cluster using different selection method
11- calculate satisfaction and dissatisfaction confident for kano features to support our diction choosing these tow categories ( one dimension and excitement )
Comment to be discussed :
Chi square can’t be used for this problem (need to check )
Adaboost and xgb can be used
This is classification problem but we use regression. how could we justify
Regression can be used
Increse depth not increse the accuracy
Make sure that weka use the right one
Categorical number round to nearest class
Depth not Necessary increase the accuracy
Classification can be used for this problem
I will repeat the following experiments 3 times
• Cross validation 5-fold
• Cross validation 10-fold
• Split to test and train
• Grid Search sklearn for selecting the best parameter
Experiments number Titel Page
1 All kano Select 13-19
2 common 20-25
3 union 30-35
4 Top1 + kano features 36-39
5 Top2+ kano features 40-42
6 Top3+ kano features 43-45
7 Male all 48-50
8 Male common 51-52
9 Female all 55-57
10 Female common 58-59
11 8 Cluster
12 5 cluster
Each experiment then conclusion about the best results we achieve
This research is about prediction for customer satisfaction using feature selection compare it with all attributes accuracy
We have 2 datasets:
First kano dataset ( kano dataset aug ); we use it for kano feature selection based on categorizing feature to 5 categories as following
1- must be
2- one dimension ( performance)
3-excitement
4- indifferent
5- reverse
Feature located under 2 and 3 categories we consider them as important and we will call them kano features so we compare them with other feature selected by feature selection method to see how is the prediction with different scenario ( take the common between them r union or kano +1 or kano +2 or kano +3 or cluster to male and female or 5 cluster or 8 clusters using kano dataset then use the same records in satisfaction dataset to cluster the data
The second dataset is satisfaction dataset the target column is satisfaction level from 1 to 5 . the name of the dataset is( satisfaction dataset 2022 )
Feature can be selected using kano and compare with other feature selection methods ( anova , lasso , natural information and chi square )
Prediction use different methods ( xgboost regression, random forest , adaboost , Decision tree and multi linear regression
Experiments include prediction using -
1- all attributes
2- kano ( one dimension and excitement )features the ry are 9 features
3 - features selected
4- common between kano and feature selection
5- union between kano and feature election
6 union kano + 1 from feature selection method
7- union kano + 2 from feature selection method
8- union kano + 3 from feature selection method
9 - cluster to make and female and predict using all and common
10 - cluster to 8 clusters and compare the feature match between. All cluster using different selection method
11- - cluster to 5 clusters and compare the feature match between. All cluster using different selection method
11- calculate satisfaction and dissatisfaction confident for kano features to support our diction choosing these tow categories ( one dimension and excitement )
Comment to be discussed :
Chi square can’t be used for this problem (need to check )
Adaboost and xgb can be used
This is classification problem but we use regression. how could we justify
Regression can be used
Increse depth not increse the accuracy
Make sure that weka use the right one
Categorical number round to nearest class
Depth not Necessary increase the accuracy
Classification can be used for this problem
I will repeat the following experiments 3 times
• Cross validation 5-fold
• Cross validation 10-fold
• Split to test and train
• Grid Search sklearn for selecting the best parameter
Experiments number Titel Page
1 All kano Select 13-19
2 common 20-25
3 union 30-35
4 Top1 + kano features 36-39
5 Top2+ kano features 40-42
6 Top3+ kano features 43-45
7 Male all 48-50
8 Male common 51-52
9 Female all 55-57
10 Female common 58-59
11 8 Cluster
12 5 cluster
Each experiment then conclusion about the best results we achieve