Вы находитесь на странице: 1из 43

DATA MINING & Warehousing

Index
S.No 1. 2. 3. 4. 5. 6. . 8. #. 1&. Experiment Demonstration of preprocessing on dataset student.arff Demonstration of preprocessing on dataset labor.arff Demonstration of Association rule process on dataset contactlenses.arff using apriori algorithm Demonstration of Association rule process on dataset test.arff using apriori algorithm Demonstration of classification rule process on dataset student.arff using j48 algorithm Demonstration of classification rule process on dataset employee.arff using j48 algorithm Demonstration of classification rule process on dataset employee.arff using id3 algorithm Demonstration of classification rule process on dataset employee.arff using na!"e bayes algorithm Demonstration of clustering rule process on dataset iris.arff using simple $%means Demonstration of clustering rule process on dataset student.arff using simple $% means Page no Signature

1. Demonstration of preprocessing on dataset student.arff Aim: 'his e(periment illustrates some of the basic data preprocessing operations that can be performed using )*+A%*(plorer. 'he sample dataset used for this e(ample is the student data a"ailable in arff format. ,tep1- .oading the data. )e can load the dataset into /e$a by clic$ing on open button in preprocessing interface and selecting the appropriate file. ,tep2- 0nce the data is loaded1 /e$a /ill recogni2e the attributes and during the scan of the data /e$a /ill compute some basic strategies on each attribute. 'he left panel in the abo"e figure sho/s the list of recogni2ed attributes /hile the top panel indicates the names of the base relation or table and the current /or$ing relation 3/hich are same initially4. ,tep3-5lic$ing on an attribute in the left panel /ill sho/ the basic statistics on the attributes for the categorical attributes the fre6uency of each attribute "alue is sho/n1 /hile for continuous attributes /e can obtain min1 ma(1 mean1 standard de"iation and de"iation etc.1 ,tep4-'he "isuali2ation in the right button panel in the form of cross%tabulation across t/o attributes. Note:/e can select another attribute using the dropdo/n list. ,tep5-,electing or filtering attributes 7emo"ing an attribute%)hen /e need to remo"e an attribute1/e can do this by using the attribute filters in /e$a.8n the filter model panel1clic$ on choose button1'his /ill sho/ a popup /indo/ /ith a list of a"ailable filters. ,croll do/n the list and select the 9/e$a.filters.unsuper"ised.attribute.remo"e: filters. ,tep 6-a4;e(t clic$ the te(tbo( immediately to the right of the choose button.8n the resulting dialog bo( enter the inde( of the attribute to be filtered out. b4<a$e sure that in"ert selection option is set to false.'he clic$ 0+ no/ in the filter bo(.you /ill see 97emo"e%7% :. c45lic$ the apply button to apply filter to this data.'his /ill remo"e the attribute and create ne/ /or$ing relation. d4,a"e the ne/ /or$ing relation as an arff file by clic$ing sa"e button on the top3button4panel.3student.arff4

Discretization 14,ometimes association rule mining can only be performed on categorical data.'his re6uires performing discreti2ation on numeric or continuous attributes.8n the follo/ing e(ample let us discreti2e age attribute. .et us di"ide the "alues of age attribute into three bins3inter"als4. =irst load the dataset into /e$a3student.arff4 ,elect the age attribute. Acti"ate filter%dialog bo( and select 9)*+A.filters.unsuper"ised.attribute.discreti2e:from the list. 'o change the defaults for the filters1clic$ on the bo( immediately to the right of the choose button. )e enter the inde( for the attribute to be discreti2ed.8n this case the attribute is age.,o /e must enter >1? corresponding to the age attribute. *nter >3? as the number of bins..ea"e the remaining field "alues as they are. 5lic$ 0+ button. 5lic$ apply in the filter panel.'his /ill result in a ne/ /or$ing relation /ith the selected attribute partition into 3 bins. ,a"e the ne/ /or$ing relation in a file called student%data%discreti2ed.arff

Dataset student .arff @relation student @attribute age AB3&13&%4&1C4&D @attribute income Alo/1 medium1 highD @attribute student Ayes1 noD @attribute credit%rating Afair1 e(cellentD @attribute buyspc Ayes1 noD @data E

B3&1 high1 no1 fair1 no B3&1 high1 no1 e(cellent1 no 3&%4&1 high1 no1 fair1 yes C4&1 medium1 no1 fair1 yes C4&1 lo/1 yes1 fair1 yes C4&1 lo/1 yes1 e(cellent1 no 3&%4&1 lo/1 yes1 e(cellent1 yes B3&1 medium1 no1 fair1 no B3&1 lo/1 yes1 fair1 no C4&1 medium1 yes1 fair1 yes B3&1 medium1 yes1 e(cellent1 yes 3&%4&1 medium1 no1 e(cellent1 yes 3&%4&1 high1 yes1 fair1 yes C4&1 medium1 no1 e(cellent1 no E

'he follo/ing screenshot sho/s the effect of discreti2ation.

2. Demonstration of preprocessing on dataset labor.arff Aim: 'his e(periment illustrates some of the basic data preprocessing operations that can be performed using )*+A%*(plorer. 'he sample dataset used for this e(ample is the labor data a"ailable in arff format. ,tep1-.oading the data. )e can load the dataset into /e$a by clic$ing on open button in preprocessing interface and selecting the appropriate file. ,tep2-0nce the data is loaded1 /e$a /ill recogni2e the attributes and during the scan of the data /e$a /ill compute some basic strategies on each attribute. 'he left panel in the abo"e figure sho/s the list of recogni2ed attributes /hile the top panel indicates the names of the base relation or table and the current /or$ing relation 3/hich are same initially4. ,tep3-5lic$ing on an attribute in the left panel /ill sho/ the basic statistics on the attributes for the categorical attributes the fre6uency of each attribute "alue is sho/n1 /hile for continuous attributes /e can obtain min1 ma(1 mean1 standard de"iation and de"iation etc.1 ,tep4-'he "isuali2ation in the right button panel in the form of cross%tabulation across t/o attributes. Note:/e can select another attribute using the dropdo/n list. ,tep5-,electing or filtering attributes 7emo"ing an attribute%)hen /e need to remo"e an attribute1/e can do this by using the attribute filters in /e$a.8n the filter model panel1clic$ on choose button1'his /ill sho/ a popup /indo/ /ith a list of a"ailable filters. ,croll do/n the list and select the 9/e$a.filters.unsuper"ised.attribute.remo"e: filters. ,tep 6-a4;e(t clic$ the te(tbo( immediately to the right of the choose button.8n the resulting dialog bo( enter the inde( of the attribute to be filtered out. b4<a$e sure that in"ert selection option is set to false.'he clic$ 0+ no/ in the filter bo(.you /ill see 97emo"e%7% :. c45lic$ the apply button to apply filter to this data.'his /ill remo"e the attribute and create ne/ /or$ing relation. d4,a"e the ne/ /or$ing relation as an arff file by clic$ing sa"e button on the top3button4panel.3labor.arff4

Discretization 14,ometimes association rule mining can only be performed on categorical data.'his re6uires performing discreti2ation on numeric or continuous attributes.8n the follo/ing e(ample let us discreti2e duration attribute. .et us di"ide the "alues of duration attribute into three bins3inter"als4. =irst load the dataset into /e$a3labor.arff4 ,elect the duration attribute. Acti"ate filter%dialog bo( and select 9)*+A.filters.unsuper"ised.attribute.discreti2e:from the list. 'o change the defaults for the filters1clic$ on the bo( immediately to the right of the choose button. )e enter the inde( for the attribute to be discreti2ed.8n this case the attribute is duration ,o /e must enter >1? corresponding to the duration attribute. *nter >1? as the number of bins..ea"e the remaining field "alues as they are. 5lic$ 0+ button. 5lic$ apply in the filter panel.'his /ill result in a ne/ /or$ing relation /ith the selected attribute partition into 1 bin. ,a"e the ne/ /or$ing relation in a file called labor%data%discreti2ed.arff Dataset labor.arff

'he follo/ing screenshot sho/s the effect of discreti2ation

. Demonstration of Association rule process on dataset contactlenses.arff using apriori algorit!m Aim: 'his e(periment illustrates some of the basic elements of asscociation rule mining using )*+A. 'he sample dataset used for this e(ample is contactlenses.arff ,tep1- 0pen the data file in )e$a *(plorer. 8t is presumed that the re6uired data fields ha"e been discreti2ed. 8n this e(ample it is age attribute. ,tep2- 5lic$ing on the associate tab /ill bring up the interface for association rule algorithm. ,tep3- )e /ill use apriori algorithm. 'his is the default algorithm. ,tep4- 8norder to change the parameters for the run 3e(ample support1 confidence etc4 /e clic$ on the te(t bo( immediately to the right of the choose button. Dataset contactlenses.arff

'he follo/ing screenshot sho/s the association rules that /ere generated /hen apriori algorithm is applied on the gi"en dataset.

". Demonstration of Association rule process on dataset test.arff using apriori algorit!m Aim: 'his e(periment illustrates some of the basic elements of asscociation rule mining using )*+A. 'he sample dataset used for this e(ample is test.arff ,tep1- 0pen the data file in )e$a *(plorer. 8t is presumed that the re6uired data fields ha"e been discreti2ed. 8n this e(ample it is age attribute. ,tep2- 5lic$ing on the associate tab /ill bring up the interface for association rule algorithm. ,tep3- )e /ill use apriori algorithm. 'his is the default algorithm. ,tep4- 8norder to change the parameters for the run 3e(ample support1 confidence etc4 /e clic$ on the te(t bo( immediately to the right of the choose button. Dataset test.arff @relation test @attribute admissionyear A2&&512&&612&& 12&&812&&#12&1&D @attribute course Acse1mech1it1eceD @data E 2&&51 cse 2&&51 it 2&&51 cse 2&&61 mech 2&&61 it 2&&61 ece 2&& 1 it 2&& 1 cse 2&&81 it 2&&81 cse 2&&#1 it

2&&#1 ece E

'he follo/ing screenshot sho/s the association rules that /ere generated /hen apriori algorithm is applied on the gi"en dataset.

#. Demonstration of classification rule process on dataset student.arff using $"% algorit!m Aim: 'his e(periment illustrates the use of j%48 classifier in /e$a. 'he sample data set used in this e(periment is 9student: data a"ailable at arff format. 'his document assumes that appropriate data pre processing has been performed. ,teps in"ol"ed in this e(periment,tep%1- )e begin the e(periment by loading the data 3student.arff4into /e$a. ,tep2- ;e(t /e select the 9classify: tab and clic$ 9choose: button t o select the 9j48:classifier. ,tep3- ;o/ /e specify the "arious parameters. 'hese can be specified by clic$ing in the te(t bo( to the right of the chose button. 8n this e(ample1 /e accept the default "alues. 'he default "ersion does perform some pruning but does not perform error pruning. ,tep4- Fnder the 9te(t: options in the main panel. )e select the 1&%fold cross "alidation as our e"aluation approach. ,ince /e don?t ha"e separate e"aluation data set1 this is necessary to get a reasonable idea of accuracy of generated model. ,tep%5- )e no/ clic$ :start: to generate the model .the Ascii "ersion of the tree as /ell as e"aluation statistic /ill appear in the right panel /hen the model construction is complete. ,tep%6- ;ote that the classification accuracy of model is about 6#E.this indicates that /e may find more /or$. 3*ither in preprocessing or in selecting current parameters for the classification4 ,tep% - ;o/ /e$a also lets us a "ie/ a graphical "ersion of the classification tree. 'his can be done by right clic$ing the last result set and selecting 9"isuali2e tree: from the pop%up menu. ,tep%8- )e /ill use our model to classify the ne/ instances. ,tep%#- 8n the main panel under 9te(t: options clic$ the 9supplied test set: radio button and then clic$ the 9set: button. 'his /ills pop%up a /indo/ /hich /ill allo/ you to open the file containing test instances.

Dataset student .arff @relation student @attribute age AB3&13&%4&1C4&D @attribute income Alo/1 medium1 highD @attribute student Ayes1 noD @attribute credit%rating Afair1 e(cellentD @attribute buyspc Ayes1 noD @data E B3&1 high1 no1 fair1 no B3&1 high1 no1 e(cellent1 no 3&%4&1 high1 no1 fair1 yes C4&1 medium1 no1 fair1 yes C4&1 lo/1 yes1 fair1 yes C4&1 lo/1 yes1 e(cellent1 no 3&%4&1 lo/1 yes1 e(cellent1 yes B3&1 medium1 no1 fair1 no B3&1 lo/1 yes1 fair1 no C4&1 medium1 yes1 fair1 yes B3&1 medium1 yes1 e(cellent1 yes 3&%4&1 medium1 no1 e(cellent1 yes 3&%4&1 high1 yes1 fair1 yes C4&1 medium1 no1 e(cellent1 no E

'he follo/ing screenshot sho/s the classification rules that /ere generated /hen j48 algorithm is applied on the gi"en dataset.

&. Demonstration of classification rule process on dataset emplo'ee.arff using $"% algorit!m Aim- 'his e(periment illustrates the use of j%48 classifier in /e$a.the sample data set used in this e(periment is 9employee:data a"ailable at arff format. 'his document assumes that appropriate data pre processing has been performed. ,teps in"ol"ed in this e(periment,tep 1- )e begin the e(periment by loading the data 3employee.arff4 into /e$a. ,tep2- ;e(t /e select the 9classify: tab and clic$ 9choose: button to select the 9j48:classifier. ,tep3- ;o/ /e specify the "arious parameters. 'hese can be specified by clic$ing in the te(t bo( to the right of the chose button. 8n this e(ample1 /e accept the default "alues the default "ersion does perform some pruning but does not perform error pruning. ,tep4- Fnder the 9te(t 9options in the main panel. )e select the 1&%fold cross "alidation as our e"aluation approach. ,ince /e don?t ha"e separate e"aluation data set1 this is necessary to get a reasonable idea of accuracy of generated model. ,tep%5- )e no/ clic$ :start: to generate the model .the A,588 "ersion of the tree as /ell as e"aluation statistic /ill appear in the right panel /hen the model construction is complete. ,tep%6- ;ote that the classification accuracy of model is about 6#E.this indicates that /e may find more /or$. 3*ither in preprocessing or in selecting current parameters for the classification4 ,tep% - ;o/ /e$a also lets us a "ie/ a graphical "ersion of the classification tree. 'his can be done by right clic$ing the last result set and selecting 9"isuali2e tree: from the pop%up menu. ,tep%8- )e /ill use our model to classify the ne/ instances. ,tep%#- 8n the main panel under 9te(t 9options clic$ the 9supplied test set: radio button and then clic$ the 9set: button. 'his /ills pop%up a /indo/ /hich /ill allo/ you to open the file containing test instances.

Data set emplo'ee.arff: @relation employee @attribute age A251 2 1 281 2#1 3&1 351 48D @attribute salaryA1&$115$11 $12&$125$13&$135$132$D @attribute performance Agood1 a"g1 poorD @data E 251 1&$1 poor 2 1 15$1 poor 2 1 1 $1 poor 281 1 $1 poor 2#1 2&$1 a"g 3&1 25$1 a"g 2#1 25$1 a"g 3&1 2&$1 a"g 351 32$1 good 481 34$1 good 481 32$1good E

'he follo/ing screenshot sho/s the classification rules that /ere generated /henj48 algorithm is applied on the gi"en dataset.

(. Demonstration of classification rule process on dataset emplo'ee.arff using id algorit!m Aim: 'his e(periment illustrates the use of id3 classifier in /e$a. 'he sample data set used in this e(periment is 9employee:data a"ailable at arff format. 'his document assumes that appropriate data pre processing has been performed. ,teps in"ol"ed in this e(periment1. )e begin the e(periment by loading the data 3employee.arff4 into /e$a. ,tep2- ne(t /e select the 9classify: tab and clic$ 9choose: button to select the 9id3:classifier. ,tep3- no/ /e specify the "arious parameters. 'hese can be specified by clic$ing in the te(t bo( to the right of the chose button. 8n this e(ample1 /e accept the default "alues his default "ersion does perform some pruning but does not perform error pruning. ,tep4- under the 9te(t 9options in the main panel. )e select the 1&%fold cross "alidation as our e"aluation approach. ,ince /e don?t ha"e separate e"aluation data set1 this is necessary to get a reasonable idea of accuracy of generated model. ,tep%5- /e no/ clic$:start:to generate the model .the A,588 "ersion of the tree as /ell as e"aluation statistic /ill appear in the right panel /hen the model construction is complete. ,tep%6- note that the classification accuracy of model is about 6#E.this indicates that /e may find more /or$. 3*ither in preprocessing or in selecting current parameters for the classification4 ,tep% - no/ /e$a also lets us a "ie/ a graphical "ersion of the classification tree. 'his can be done by right clic$ing the last result set and selecting 9"isuali2e tree: from the pop%up menu. ,tep%8- /e /ill use our model to classify the ne/ instances. ,tep%#- 8n the main panel under 9te(t 9options clic$ the 9supplied test set: radio button and then clic$ the 9set: button. 'his /ill sho/ pop%up /indo/ /hich /ill allo/ you to open the file containing test instances.

Data set emplo'ee.arff: @relation employee @attribute age A251 2 1 281 2#1 3&1 351 48D @attribute salaryA1&$115$11 $12&$125$13&$135$132$D @attribute performance Agood1 a"g1 poorD @data E 251 1&$1 poor 2 1 15$1 poor 2 1 1 $1 poor 281 1 $1 poor 2#1 2&$1 a"g 3&1 25$1 a"g 2#1 25$1 a"g 3&1 2&$1 a"g 351 32$1 good 481 34$1 good 481 32$1 good E

'he follo/ing screenshot sho/s the classification rules that /ere generated /hen id3 algorithm is applied on the gi"en dataset.

%.Demonstration of classification rule process on dataset emplo'ee.arff using na)*e ba'es algorit!m Aim: 'his e(periment illustrates the use of na!"e bayes classifier in /e$a. 'he sample data set used in this e(periment is 9employee:data a"ailable at arff format. 'his document assumes that appropriate data pre processing has been performed. ,teps in"ol"ed in this e(periment1. )e begin the e(periment by loading the data 3employee.arff4 into /e$a. ,tep2- ne(t /e select the 9classify: tab and clic$ 9choose: button to select the 9id3:classifier. ,tep3- no/ /e specify the "arious parameters. 'hese can be specified by clic$ing in the te(t bo( to the right of the chose button. 8n this e(ample1 /e accept the default "alues his default "ersion does perform some pruning but does not perform error pruning. ,tep4- under the 9te(t 9options in the main panel. )e select the 1&%fold cross "alidation as our e"aluation approach. ,ince /e don?t ha"e separate e"aluation data set1 this is necessary to get a reasonable idea of accuracy of generated model. ,tep%5- /e no/ clic$:start:to generate the model .the A,588 "ersion of the tree as /ell as e"aluation statistic /ill appear in the right panel /hen the model construction is complete. ,tep%6- note that the classification accuracy of model is about 6#E.this indicates that /e may find more /or$. 3*ither in preprocessing or in selecting current parameters for the classification4 ,tep% - no/ /e$a also lets us a "ie/ a graphical "ersion of the classification tree. 'his can be done by right clic$ing the last result set and selecting 9"isuali2e tree: from the pop%up menu. ,tep%8- /e /ill use our model to classify the ne/ instances. ,tep%#- 8n the main panel under 9te(t 9options clic$ the 9supplied test set: radio button and then clic$ the 9set: button. 'his /ill sho/ pop%up /indo/ /hich /ill allo/ you to open the file containing test instances.

Data set emplo'ee.arff: @relation employee @attribute age A251 2 1 281 2#1 3&1 351 48D @attribute salaryA1&$115$11 $12&$125$13&$135$132$D @attribute performance Agood1 a"g1 poorD @data E 251 1&$1 poor 2 1 15$1 poor 2 1 1 $1 poor 281 1 $1 poor 2#1 2&$1 a"g 3&1 25$1 a"g 2#1 25$1 a"g 3&1 2&$1 a"g 351 32$1 good 481 34$1 good 481 32$1 good E

'he follo/ing screenshot sho/s the classification rules that /ere generated /hen nai"e bayes algorithm is applied on the gi"en dataset.

+. Demonstration of clustering rule process on dataset iris.arff using simple ,-means Aim- 'his e(periment illustrates the use of simple $%mean clustering /ith )e$a e(plorer. 'he sample data set used for this e(ample is based on the iris data a"ailable in A7== format. 'his document assumes that appropriate preprocessing has been performed. 'his iris dataset includes 15& instances. Steps in*ol*ed in t!is Experiment ,tep 1- 7un the )e$a e(plorer and load the data file iris.arff in preprocessing interface. ,tep 2- 8norder to perform clustering select the >cluster? tab in the e(plorer and clic$ on the choose button. 'his step results in a dropdo/n list of a"ailable clustering algorithms. ,tep 3 - 8n this case /e select >simple $%means?. ,tep 4- ;e(t clic$ in te(t button to the right of the choose button to get popup /indo/ sho/n in the screenshots. 8n this /indo/ /e enter si( on the number of clusters and /e lea"e the "alue of the seed on as it is. 'he seed "alue is used in generating a random number /hich is used for ma$ing the internal assignments of instances of clusters. ,tep 5 - 0nce of the option ha"e been specified. )e run the clustering algorithm there /e must ma$e sure that they are in the >cluster mode? panel. 'he use of training set option is selected and then /e clic$ >start? button. 'his process and resulting /indo/ are sho/n in the follo/ing screenshots. ,tep 6 - 'he result /indo/ sho/s the centroid of each cluster as /ell as statistics on the number and the percent of instances assigned to different clusters. Gere clusters centroid are means "ectors for each clusters. 'his clusters can be used to characteri2ed the cluster.=or eg1 the centroid of cluster1 sho/s the class iris."ersicolor mean "alue of the sepal length is 5.4 &61 sepal /idth 2.4 651 petal /idth 1.12#41 petal length 3. #41. ,tep - Another /ay of understanding characterstics of each cluster through "isuali2ation 1/e can do this1 try right clic$ing the result set on the result. .ist panel and selecting the "isuali2e cluster assignments.

'he follo/ing screenshot sho/s the clustering rules that /ere generated /hen simple $ means algorithm is applied on the gi"en dataset.

Interpretation of t!e abo*e *isualization =rom the abo"e "isuali2ation1 /e can understand the distribution of sepal length and petal length in each cluster. =or instance1 for each cluster is dominated by petal length. 8n this case by changing the color dimension to other attributes /e can see their distribution /ith in each of the cluster. ,tep 8- )e can assure that resulting dataset /hich included each instance along /ith its assign cluster. 'o do so /e clic$ the sa"e button in the "isuali2ation /indo/ and sa"e the result iris $%mean .'he top portion of this file is sho/n in the follo/ing figure.

1.. Demonstration of clustering rule process on dataset student.arff using simple ,means Aim- 'his e(periment illustrates the use of simple $%mean clustering /ith )e$a e(plorer. 'he sample data set used for this e(ample is based on the student data a"ailable in A7== format. 'his document assumes that appropriate preprocessing has been performed. 'his istudent dataset includes 14 instances. ,teps in"ol"ed in this *(periment ,tep 1- 7un the )e$a e(plorer and load the data file student.arff in preprocessing interface. ,tep 2- 8norder to perform clustering select the >cluster? tab in the e(plorer and clic$ on the choose button. 'his step results in a dropdo/n list of a"ailable clustering algorithms. ,tep 3 - 8n this case /e select >simple $%means?. ,tep 4- ;e(t clic$ in te(t button to the right of the choose button to get popup /indo/ sho/n in the screenshots. 8n this /indo/ /e enter si( on the number of clusters and /e lea"e the "alue of the seed on as it is. 'he seed "alue is used in generating a random number /hich is used for ma$ing the internal assignments of instances of clusters. ,tep 5 - 0nce of the option ha"e been specified. )e run the clustering algorithm there /e must ma$e sure that they are in the >cluster mode? panel. 'he use of training set option is selected and then /e clic$ >start? button. 'his process and resulting /indo/ are sho/n in the follo/ing screenshots. ,tep 6 - 'he result /indo/ sho/s the centroid of each cluster as /ell as statistics on the number and the percent of instances assigned to different clusters. Gere clusters centroid are means "ectors for each clusters. 'his clusters can be used to characteri2ed the cluster. ,tep - Another /ay of understanding characterstics of each cluster through "isuali2ation 1/e can do this1 try right clic$ing the result set on the result. .ist panel and selecting the "isuali2e cluster assignments. Interpretation of t!e abo*e *isualization =rom the abo"e "isuali2ation1 /e can understand the distribution of age and instance number in each cluster. =or instance1 for each cluster is dominated by age. 8n this case by changing the color dimension to other attributes /e can see their distribution /ith in each of the cluster. ,tep 8- )e can assure that resulting dataset /hich included each instance along /ith its assign cluster. 'o do so /e clic$ the sa"e button in the "isuali2ation /indo/ and sa"e the result student $%mean .'he top portion of this file is sho/n in the follo/ing figure.

Dataset student .arff @relation student @attribute age AB3&13&%4&1C4&D @attribute income Alo/1medium1highD @attribute student Ayes1noD @attribute credit%rating Afair1e(cellentD @attribute buyspc Ayes1noD @data E B3&1 high1 no1 fair1 no B3&1 high1 no1 e(cellent1 no 3&%4&1 high1 no1 fair1 yes C4&1 medium1 no1 fair1 yes C4&1 lo/1 yes1 fair1 yes C4&1 lo/1 yes1 e(cellent1 no 3&%4&1 lo/1 yes1 e(cellent1 yes B3&1 medium1 no1 fair1 no B3&1 lo/1 yes1 fair1 no C4&1 medium1 yes1 fair1 yes B3&1 medium1 yes1 e(cellent1 yes 3&%4&1 medium1 no1 e(cellent1 yes 3&%4&1 high1 yes1 fair1 yes C4&1 medium1 no1 e(cellent1 no E

'he follo/ing screenshot sho/s the clustering rules that /ere generated /hen simple $% means algorithm is applied on the gi"en dataset.

Вам также может понравиться