"Bi thing" data mining algorithms--verification of accuracy

Accuracy Validation Example 1 :--based on kingdoms Database

Data preparation:

Mining Model:
In order: Naive Bayes algorithm, cluster analysis algorithm, decision tree algorithm, neural network algorithm, logistic regression algorithm, correlation algorithm

Lift Chart:

Rank to:
1. Neural Network algorithm (92.69% 0.99)
2. Logistic regression algorithm (92.39% 0.99)
3. Decision Tree Algorithm (91.19% 0.98)
4. Correlation algorithm (90.6% 0.98)
5. Clustering analysis Algorithm (89.25% 0.96)
6. Naive Bayes algorithm (87.61 0.96)

Naive Bayes algorithm--Classification matrix

Description
The 538 samples of other classes had 482 predictions correctly, 32 were divided into military advisers, 24 were divided into generals, and the prediction accuracy was 89.59%.
20 samples of the adviser have 13 predictions correctly, 7 are divided into other classes, the prediction accuracy rate is 65%;
The general 112 samples had 92 predictions correctly, 16 were divided into other classes, 4 were divided into advisers, and the prediction accuracy was 82.14%.
Clustering analysis algorithm--Classification matrix

Description
The 538 samples of other classes had 536 predictions correctly, 2 were divided into general categories, and the prediction accuracy was 99.63%.
20 samples of the adviser have 0 predictions correctly, 20 are divided into other classes, the prediction accuracy rate is 0%;
The general 112 samples had 62 predictions correctly, 50 were divided into other classes, and the prediction accuracy rate was 55.36%.
Decision Tree algorithm--Classification matrix

Description
The 538 samples of the other classes had 538 predictions correctly, and the prediction accuracy was 100%;
20 samples of the adviser have 0 predictions correctly, 20 are divided into other classes, the prediction accuracy rate is 0%;
The general 112 samples had 73 predictions correctly, 39 were divided into other classes, and the prediction accuracy rate was 65.18%.
Neural Network algorithm--classification matrix

Description
The 538 samples of other classes had 524 predictions correctly, 5 were divided into military advisers, 9 were divided into generals, and the prediction accuracy was 97.4%.
20 samples of the adviser have 5 predictions correctly, 15 are divided into other classes, the prediction accuracy rate is 25%;
The general 112 samples had 92 predictions correctly, 20 were divided into other classes, and the prediction accuracy rate was 82.14%.
Logistic regression algorithm--Classification matrix

Description
The 538 samples of other classes had 526 predictions correctly, 6 were divided into military advisers, 6 were divided into generals, and the prediction accuracy was 97.77%.
20 samples of the adviser have 5 predictions correctly, 15 are divided into other classes, the prediction accuracy rate is 25%;
The general 112 samples had 88 predictions correctly, 24 were divided into other classes, and the prediction accuracy rate was 78.57%.
Correlation algorithm--classification matrix

Description
The 538 samples of other classes had 519 predictions correctly, 19 were divided into advisers, and the prediction accuracy was 96.47%.
20 samples of the adviser have 0 predictions correctly, 20 are divided into other classes, the prediction accuracy rate is 0%;
The general 112 samples had 88 predictions correctly, 24 were divided into other classes, and the prediction accuracy rate was 78.57%.
Classification Matrix--prediction of correct rate summary analysis:

	Other	Consigliere	General
Neural Network algorithm	97.4%	25%	82.14%
Logistic regression algorithm	97.77%	25%	78.57%
Decision Tree algorithm	100%	0%	65.18%
Correlation algorithm	96.47%	0%	78.57%
Clustering Analysis algorithm	99.63%	0%	55.36%
Naive Bayes algorithm	89.59%	65%	82.14%

It can be seen that the naive Bayes algorithm has the highest accuracy in predicting the status of a strategist, reaching 65%, decision Tree Algorithm, association algorithm, clustering algorithm 0%, neural network algorithm and logistic regression algorithm 25%;
The decision tree algorithm has the highest accuracy in predicting other identities, reaching 100%.
Neural network algorithm and Naive Bayes algorithm are used to predict the accuracy of general identity, and reach 82.14%.

Accuracy Verification example 2:--based on stock data
Data preparation:

The mining model is in turn:
Stockclustering Clustering Analysis algorithm
Strockdecisiontrees Decision Tree Algorithm
Stockneuralnetwork Neural Network algorithm
Stocklogistic Logistic regression algorithm

Lift Chart:

rank to:
1. Logistic regression algorithm (49.73% 0.52)
2. Neural network algorithm (49.63% 0.53)
3. Clustering analysis Algorithm (48.13% 0.51)
4. Decision Tree Algorithm (47.28% 0.50)
Cluster analysis algorithm--Classification matrix:

Description
The flat 114 samples had 0 predictions correctly, 91 were divided into fall, 23 were divided into up, the prediction accuracy was 0%;
The 443 samples of the fall have 340 predictions correctly, 103 errors are up, and the forecast accuracy rate is 76.75%;
The 380-up sample has 111 predictions correctly, 269 errors fall, and the forecast accuracy rate is 29.21%.

Decision Tree algorithm--Classification matrix:

Description
The flat 114 samples have 0 predictions correctly, 114 mistakes fall, and the forecast accuracy rate is 0%;
443 of the 443 samples dropped were correctly predicted and the correct rate was 100%;
The 380-up sample has 0 predictions correctly, 380 errors fall, and the forecast accuracy rate is 0%.

Neural Network algorithm--Classification matrix:

Description
The flat 114 samples had 0 predictions correctly, 60 were divided into fall, 54 were divided into up, the prediction accuracy was 0%;
The 443 samples of the fall have 277 predictions correctly, 166 errors are up, and the forecast accuracy rate is 62.53%;
The 380-up sample has 188 predictions correctly, 192 errors fall, and the forecast accuracy rate is 49.47%.

Logistic regression algorithm--Classification matrix:

Description
The flat 114 samples had 0 predictions correctly, 89 were divided into fall, 25 were divided into up, the prediction accuracy was 0%;
The 443 samples of the fall have 380 predictions correctly, 63 errors are up, and the forecast accuracy rate is 85.78%;
The 380-up sample has 86 predictions correctly, 294 errors fall, and the forecast accuracy rate is 22.63%.

Classification Matrix--prediction of correct rate summary analysis:

	Flat	Fell	Rose
Logistic regression algorithm	0%	85.78%	22.63%
Neural Network algorithm	0%	62.53%	49.47%
Clustering Analysis algorithm	0%	76.75%	29.21%
Decision Tree algorithm	0%	100%	0%

"Bi thing" data mining algorithms--verification of accuracy