Intel DAAL AI加速 ——传统决策树和随机森林

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
# file: dt_cls_dense_batch.py
#===============================================================================
# Copyright 2014-2018 Intel Corporation.
#
# This software and the related documents are Intel copyrighted  materials,  and
# your use of  them is  governed by the  express license  under which  they were
# provided to you (License).  Unless the License provides otherwise, you may not
# use, modify, copy, publish, distribute,  disclose or transmit this software or
# the related documents without Intel's prior written permission.
#
# This software and the related documents  are provided as  is,  with no express
# or implied  warranties,  other  than those  that are  expressly stated  in the
# License.
#===============================================================================
 
## <a name="DAAL-EXAMPLE-PY-DT_CLS_DENSE_BATCH"></a>
## \example dt_cls_dense_batch.py
 
import os
import sys
 
from daal.algorithms.decision_tree.classification import prediction, training
from daal.algorithms import classifier
from daal.data_management import (
    FileDataSource, DataSourceIface, NumericTableIface, HomogenNumericTable, MergedNumericTable
)
utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
if utils_folder not in sys.path:
    sys.path.insert(0, utils_folder)
from utils import printNumericTables
 
DAAL_PREFIX = os.path.join('..', 'data')
 
# Input data set parameters
trainDatasetFileName = os.path.join(DAAL_PREFIX, 'batch', 'decision_tree_train.csv')
pruneDatasetFileName = os.path.join(DAAL_PREFIX, 'batch', 'decision_tree_prune.csv')
testDatasetFileName = os.path.join(DAAL_PREFIX, 'batch', 'decision_tree_test.csv')
 
nFeatures = 5
nClasses = 5
 
# Model object for the decision tree classification algorithm
model = None
predictionResult = None
testGroundTruth = None
 
 
def trainModel():
    global model
 
    # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
    trainDataSource = FileDataSource(
        trainDatasetFileName,
        DataSourceIface.notAllocateNumericTable,
        DataSourceIface.doDictionaryFromContext
    )
 
    # Create Numeric Tables for training data and labels
    trainData = HomogenNumericTable(nFeatures, 0, NumericTableIface.notAllocate)
    trainGroundTruth = HomogenNumericTable(1, 0, NumericTableIface.notAllocate)
    mergedData = MergedNumericTable(trainData, trainGroundTruth)
 
    # Retrieve the data from the input file
    trainDataSource.loadDataBlock(mergedData)
 
    # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
    pruneDataSource = FileDataSource(
        pruneDatasetFileName,
        DataSourceIface.notAllocateNumericTable,
        DataSourceIface.doDictionaryFromContext
    )
 
    # Create Numeric Tables for pruning data and labels
    pruneData = HomogenNumericTable(nFeatures, 0, NumericTableIface.notAllocate)
    pruneGroundTruth = HomogenNumericTable(1, 0, NumericTableIface.notAllocate)
    pruneMergedData = MergedNumericTable(pruneData, pruneGroundTruth)
 
    # Retrieve the data from the input file
    pruneDataSource.loadDataBlock(pruneMergedData)
 
    # Create an algorithm object to train the decision tree classification model
    algorithm = training.Batch(nClasses)
 
    # Pass the training data set and dependent values to the algorithm
    algorithm.input.set(classifier.training.data, trainData)
    algorithm.input.set(classifier.training.labels, trainGroundTruth)
    algorithm.input.setTable(training.dataForPruning, pruneData)
    algorithm.input.setTable(training.labelsForPruning, pruneGroundTruth)
 
    # Train the decision tree classification model and retrieve the results of the training algorithm
    trainingResult = algorithm.compute()
    model = trainingResult.get(classifier.training.model)
 
def testModel():
    global testGroundTruth, predictionResult
 
    # Initialize FileDataSource<CSVFeatureManager> to retrieve the test data from a .csv file
    testDataSource = FileDataSource(
        testDatasetFileName,
        DataSourceIface.notAllocateNumericTable,
        DataSourceIface.doDictionaryFromContext
    )
 
    # Create Numeric Tables for testing data and labels
    testData = HomogenNumericTable(nFeatures, 0, NumericTableIface.notAllocate)
    testGroundTruth = HomogenNumericTable(1, 0, NumericTableIface.notAllocate)
    mergedData = MergedNumericTable(testData, testGroundTruth)
 
    # Retrieve the data from input file
    testDataSource.loadDataBlock(mergedData)
 
    # Create algorithm objects for decision tree classification prediction with the default method
    algorithm = prediction.Batch()
 
    # Pass the testing data set and trained model to the algorithm
    #print("Number of columns: {}".format(testData.getNumberOfColumns()))
    algorithm.input.setTable(classifier.prediction.data,  testData)
    algorithm.input.setModel(classifier.prediction.model, model)
 
    # Compute prediction results and retrieve algorithm results
    # (Result class from classifier.prediction)
    predictionResult = algorithm.compute()
 
 
def printResults():
 
    printNumericTables(
        testGroundTruth,
        predictionResult.get(classifier.prediction.prediction),
        "Ground truth", "Classification results",
        "Decision tree classification results (first 20 observations):",
        20, flt64=False
    )
 
if __name__ == "__main__":
 
    trainModel()
    testModel()
    printResults()

　　

随机森林的：

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
# file: df_cls_dense_batch.py
#===============================================================================
# Copyright 2014-2018 Intel Corporation.
#
# This software and the related documents are Intel copyrighted  materials,  and
# your use of  them is  governed by the  express license  under which  they were
# provided to you (License).  Unless the License provides otherwise, you may not
# use, modify, copy, publish, distribute,  disclose or transmit this software or
# the related documents without Intel's prior written permission.
#
# This software and the related documents  are provided as  is,  with no express
# or implied  warranties,  other  than those  that are  expressly stated  in the
# License.
#===============================================================================
 
## <a name="DAAL-EXAMPLE-PY-DF_CLS_DENSE_BATCH"></a>
## \example df_cls_dense_batch.py
 
import os
import sys
 
from daal.algorithms import decision_forest
from daal.algorithms.decision_forest.classification import prediction, training
from daal.algorithms import classifier
from daal.data_management import (
    FileDataSource, DataSourceIface, NumericTableIface, HomogenNumericTable,
    MergedNumericTable, features
)
 
utils_folder = os.path.realpath(os.path.abspath(os.path.dirname(os.path.dirname(__file__))))
if utils_folder not in sys.path:
    sys.path.insert(0, utils_folder)
from utils import printNumericTable, printNumericTables
 
DAAL_PREFIX = os.path.join('..', 'data')
 
# Input data set parameters
trainDatasetFileName = os.path.join(DAAL_PREFIX, 'batch', 'df_classification_train.csv')
testDatasetFileName = os.path.join(DAAL_PREFIX, 'batch', 'df_classification_test.csv')
 
nFeatures = 3
nClasses = 5
 
# Decision forest parameters
nTrees = 10
minObservationsInLeafNode = 8
 
# Model object for the decision forest classification algorithm
model = None
predictionResult = None
testGroundTruth = None
 
 
def trainModel():
    global model
 
    # Initialize FileDataSource<CSVFeatureManager> to retrieve the input data from a .csv file
    trainDataSource = FileDataSource(
        trainDatasetFileName,
        DataSourceIface.notAllocateNumericTable,
        DataSourceIface.doDictionaryFromContext
    )
 
    # Create Numeric Tables for training data and labels
    trainData = HomogenNumericTable(nFeatures, 0, NumericTableIface.notAllocate)
    trainGroundTruth = HomogenNumericTable(1, 0, NumericTableIface.notAllocate)
    mergedData = MergedNumericTable(trainData, trainGroundTruth)
 
    # Retrieve the data from the input file
    trainDataSource.loadDataBlock(mergedData)
 
    #  Get the dictionary and update it with additional information about data
    dict = trainData.getDictionary()
 
    #  Add a feature type to the dictionary
    dict[0].featureType = features.DAAL_CONTINUOUS
    dict[1].featureType = features.DAAL_CONTINUOUS
    dict[2].featureType = features.DAAL_CATEGORICAL
 
    # Create an algorithm object to train the decision forest classification model
    algorithm = training.Batch(nClasses)
    algorithm.parameter.nTrees = nTrees
    algorithm.parameter.minObservationsInLeafNode = minObservationsInLeafNode
    algorithm.parameter.featuresPerNode = nFeatures
    algorithm.parameter.varImportance = decision_forest.training.MDI
    algorithm.parameter.resultsToCompute = decision_forest.training.computeOutOfBagError
 
    # Pass the training data set and dependent values to the algorithm
    algorithm.input.set(classifier.training.data, trainData)
    algorithm.input.set(classifier.training.labels, trainGroundTruth)
 
    # Train the decision forest classification model and retrieve the results of the training algorithm
    trainingResult = algorithm.compute()
    model = trainingResult.get(classifier.training.model)
    printNumericTable(trainingResult.getTable(training.variableImportance), "Variable importance results: ")
    printNumericTable(trainingResult.getTable(training.outOfBagError), "OOB error: ")
 
def testModel():
    global testGroundTruth, predictionResult
 
    # Initialize FileDataSource<CSVFeatureManager> to retrieve the test data from a .csv file
    testDataSource = FileDataSource(
        testDatasetFileName,
        DataSourceIface.notAllocateNumericTable,
        DataSourceIface.doDictionaryFromContext
    )
 
    # Create Numeric Tables for testing data and labels
    testData = HomogenNumericTable(nFeatures, 0, NumericTableIface.notAllocate)
    testGroundTruth = HomogenNumericTable(1, 0, NumericTableIface.notAllocate)
    mergedData = MergedNumericTable(testData, testGroundTruth)
 
    # Retrieve the data from input file
    testDataSource.loadDataBlock(mergedData)
 
    #  Get the dictionary and update it with additional information about data
    dict = testData.getDictionary()
 
    #  Add a feature type to the dictionary
    dict[0].featureType = features.DAAL_CONTINUOUS
    dict[1].featureType = features.DAAL_CONTINUOUS
    dict[2].featureType = features.DAAL_CATEGORICAL
 
    # Create algorithm objects for decision forest classification prediction with the default method
    algorithm = prediction.Batch(nClasses)
 
    # Pass the testing data set and trained model to the algorithm
    algorithm.input.setTable(classifier.prediction.data,  testData)
    algorithm.input.setModel(classifier.prediction.model, model)
 
    # Compute prediction results and retrieve algorithm results
    # (Result class from classifier.prediction)
    predictionResult = algorithm.compute()
 
 
def printResults():
    printNumericTable(predictionResult.get(classifier.prediction.prediction),"Decision forest prediction results (first 10 rows):",10)
    printNumericTable(testGroundTruth, "Ground truth (first 10 rows):", 10);
 
if __name__ == "__main__":
 
    trainModel()
    testModel()
    printResults()

　　

posted @   bonelee  阅读(915)  评论(0)  编辑  收藏  举报
编辑推荐：
· 记一次.NET内存居高不下排查解决与启示
· 探究高空视频全景AR技术的实现原理
· 理解Rust引用及其生命周期标识（上）
· 浏览器原生「磁吸」效果！Anchor Positioning 锚点定位神器解析
· 没有源码，如何修改代码逻辑？
阅读排行：
· 全程不用写代码，我用AI程序员写了一个飞机大战
· MongoDB 8.0这个新功能碉堡了，比商业数据库还牛
· 记一次.NET内存居高不下排查解决与启示
· 白话解读 Dapr 1.15：你的「微服务管家」又秀新绝活了
· DeepSeek 开源周回顾「GitHub 热点速览」
历史上的今天：
2017-09-25 Vim 删除不包含指定字符串的行及统计匹配个数
2017-09-25 pandas入门10分钟——serries其实就是data frame的一列数据
2017-09-25 转载-让PIP源使用国内镜像，提升下载速度和安装成功率
点击右上角即可分享
微信分享提示