jofunch / lt2212-v20-a2 Goto Github PK

Pt1.

the preprocessing done in part 1 includes tokenisation on whitespace, lowering capital, and removing numbers and symbols. In the code, you'll also find two additional preprocessing strings, but those I took away as they completely dropped the overall classification rate.

Of course, these could have been included, which would have improved the time overall, but the results where slightly better without.

pt 2

in the original script I used SVD where I in the bonus part used PCA.

pt 3

I used KneightborsClassifier and Naieve Bayes.

Following is the Accuracy, Precision, recall, and F-Measure of respectively KNeightborsClassifier, NaiveBayes, and DecisionTreeClassifier.

Part 4

KNeighborsClassifier:

Accuracy: 0.36074270557029176 Precision: 0.42539017448796657 recall: 0.35858746911955525 F-Measure: 0.36878891330406305

macro avg precision 0.43
weighted avg precision 0.42

GaussianNB:

Part 4

Accuracy: 0.7840848806366048 Precision: 0.7810080602806917 recall: 0.7812991445206692 F-Measure: 0.7794538461036141

micro avg 0.78
macro avg 0.79
weighted avg 0.79

Part 2 : Dimension Reduction KNeighborsClassifier:

1000 dims

Accuracy: 0.35119363395225467 Precision: 0.3944648770759483 recall: 0.34902136136974315 F-Measure: 0.35631061609521597

weighted avg 0.39 macro avg 0.39 micro avg 0.35

500 dims

Accuracy: 0.32652519893899207 Precision: 0.3638358860067592 recall: 0.3224298594295505 F-Measure: 0.32773164682021044

micro avg 0.33
macro avg 0.36
weighted avg 0.36

100

Accuracy: 0.2596816976127321 Precision: 0.28666938401614084 recall: 0.2576973034792129 F-Measure: 0.25805858738024634

micro avg 0.26
macro avg 0.29
weighted avg 0.28

50 Accuracy: 0.2116710875331565 Precision: 0.23158588668140742 recall: 0.2124250243834625 F-Measure: 0.20998849832866964

micro avg 0.21
macro avg 0.23
weighted avg 0.23

25

Accuracy: 0.18859416445623342 Precision: 0.20685958564412302 recall: 0.1850404606792142 F-Measure: 0.18518210753269518

micro avg 0.19
macro avg 0.21
weighted avg 0.21

10

Accuracy: 0.13607427055702917 Precision: 0.15513412482454467 recall: 0.13742713357849895 F-Measure: 0.13605680942491366

micro avg 0.14
macro avg 0.16 weighted avg 0.15

GaussianNB:

Part 2 : -- Dims

1000

Accuracy: 0.1506631299734748 Precision: 0.2984541330724396 recall: 0.1458518873130346 F-Measure: 0.13439574213279187

micro avg 0.15
macro avg 0.30
weighted avg 0.30

500

Accuracy: 0.15172413793103448 Precision: 0.25378986350338667 recall: 0.14640659093543346 F-Measure: 0.13325309253376155

micro avg 0.15
macro avg 0.25
weighted avg 0.26

100

Accuracy: 0.14668435013262598 Precision: 0.18623494498960608 recall: 0.1404750284848092 F-Measure: 0.1090592456641849

micro avg 0.15
macro avg 0.19
weighted avg 0.19

50

Accuracy: 0.11750663129973475 Precision: 0.17214348574387145 recall: 0.11911240608034306 F-Measure: 0.09222440820549754

micro avg 0.12
macro avg 0.17
weighted avg 0.18

25

Accuracy: 0.10954907161803713 Precision: 0.13318518561650536 recall: 0.10708751907207956 F-Measure: 0.08089462669519902

micro avg 0.11
macro avg 0.13
weighted avg 0.13

10

Accuracy: 0.10344827586206896 Precision: 0.10807676027148769 recall: 0.09842247189346343 F-Measure: 0.068969706248238

micro avg 0.10
macro avg 0.11
weighted avg 0.11

Bonus: PCA instead of SVD

KNeighborsClassifier no reduction

Accuracy: 0.356763925729443 Precision: 0.4243483125674448 recall: 0.3545013786260134 F-Measure: 0.364863111629578

micro avg 0.36
macro avg 0.42
weighted avg 0.42

1000

Accuracy: 0.3506631299734748 Precision: 0.3969953034579255 recall: 0.34732195182452763 F-Measure: 0.3528234199823011

micro avg 0.35
macro avg 0.40
weighted avg 0.40

500

Accuracy: 0.32148541114058354 Precision: 0.3563390102015113 recall: 0.3190944181617683 F-Measure: 0.3235793566570823

micro avg 0.32
macro avg 0.36
weighted avg 0.36

100

Accuracy: 0.253315649867374 Precision: 0.2837534053566474 recall: 0.25123193189365356 F-Measure: 0.25284461794548757

micro avg 0.25
macro avg 0.28
weighted avg 0.29

50

Accuracy: 0.21962864721485412 Precision: 0.24182825067551902 recall: 0.2176554602343365 F-Measure: 0.21718050560493357

micro avg 0.22
macro avg 0.24
weighted avg 0.25

25

Accuracy: 0.19018567639257294 Precision: 0.2033133454861892 recall: 0.18603215694039596 F-Measure: 0.18461765317592121

micro avg 0.19
macro avg 0.20
weighted avg 0.21

10

Accuracy: 0.14058355437665782 Precision: 0.15154943383529945 recall: 0.1384169577317447 F-Measure: 0.13639201739580056

micro avg 0.14
macro avg 0.15
weighted avg 0.15

GaussianNB

500

Accuracy: 0.16127320954907162 Precision: 0.2741685413011943 recall: 0.15230333640854354 F-Measure: 0.141600438288851

micro avg 0.16
macro avg 0.27
weighted avg 0.27

100 Accuracy: 0.1464190981432361 Precision: 0.20664946844217041 recall: 0.14451389370590378 F-Measure: 0.11925305839178409

micro avg 0.15
macro avg 0.21
weighted avg 0.21

50

Accuracy: 0.12175066312997347 Precision: 0.1562988181644729 recall: 0.11964953069489952 F-Measure: 0.09485836614068628

micro avg 0.12
macro avg 0.16
weighted avg 0.16

25

Accuracy: 0.10557029177718832 Precision: 0.1360738286438728 recall: 0.10688254301064617 F-Measure: 0.08131105078736581

micro avg 0.11
macro avg 0.14
weighted avg 0.14

10

Accuracy: 0.09363395225464191 Precision: 0.11434426880204282 recall: 0.09319618556620782 F-Measure: 0.06685692286022539

micro avg 0.09
macro avg 0.11
weighted avg 0.12