def build_sentiment(classifier, name, with_proba=True, **pmml_options):
pipeline = PMMLPipeline([
("tf-idf",
TfidfVectorizer(
analyzer="word",
preprocessor=None,
strip_accents=None,
lowercase=True,
token_pattern=None,
tokenizer=Splitter(),
stop_words="english",
ngram_range=(1, 2),
norm=None,
dtype=(numpy.float32 if isinstance(
classifier, RandomForestClassifier) else numpy.float64))),
("selector", SelectKBest(f_classif, k=500)), ("classifier", classifier)
])
pipeline.fit(sentiment_X, sentiment_y)
pipeline.configure(**pmml_options)
store_pkl(pipeline, name + ".pkl")
score = DataFrame(pipeline.predict(sentiment_X), columns=["Score"])
if with_proba == True:
score_proba = DataFrame(pipeline.predict_proba(sentiment_X),
columns=["probability(0)", "probability(1)"])
score = pandas.concat((score, score_proba), axis=1)
store_csv(score, name + ".csv")
def test_call(self):
splitter = Splitter()
self.assertEqual((), splitter(""))
self.assertEqual((), splitter("."))
self.assertEqual(("one", ), splitter("one"))
self.assertEqual(("++one", ), splitter("++one"))
self.assertEqual(("one++", ), splitter("one++"))
self.assertEqual(("one", ), splitter("--one"))
self.assertEqual(("one", ), splitter("one--"))
self.assertEqual(("one", "two", "three"), splitter("one two three"))
self.assertEqual(("one", "t,w.o", "three"),
splitter(",one _t,w.o_ three."))
("densifier", DenseTransformer()),
("selector", SelectKBest(f_classif, k = 500)),
("classifier", classifier)
])
pipeline.fit(sentiment_X, sentiment_y)
pipeline.configure(**pmml_options)
store_pmml(pipeline, name)
score = DataFrame(pipeline.predict(sentiment_X), columns = ["Score"])
if with_proba:
score_proba = DataFrame(pipeline.predict_proba(sentiment_X), columns = ["probability(0)", "probability(1)"])
score = pandas.concat((score, score_proba), axis = 1)
store_csv(score, name)
if "Sentiment" in datasets:
pmml_textindex_args = dict(analyzer = "word", preprocessor = None, strip_accents = None, dtype = numpy.float64)
build_sentiment(LinearDiscriminantAnalysis(), TfidfVectorizer(tokenizer = Splitter(), ngram_range = (1, 3), norm = None, **pmml_textindex_args), "LinearDiscriminantAnalysisSentiment")
build_sentiment(LinearSVC(random_state = 13), CountVectorizer(tokenizer = Splitter(), ngram_range = (1, 2), **pmml_textindex_args), "LinearSVCSentiment", with_proba = False)
build_sentiment(LogisticRegression(multi_class = "ovr"), TfidfVectorizer(stop_words = "english", tokenizer = Matcher(), ngram_range = (1, 3), binary = True, norm = None, **pmml_textindex_args), "LogisticRegressionSentiment")
build_sentiment(RandomForestClassifier(max_depth = 8, min_samples_leaf = 10, n_estimators = 31, random_state = 13), CountVectorizer(ngram_range = (1, 2), **pmml_textindex_args), "RandomForestSentiment")
build_sentiment(XGBClassifier(objective = "binary:logistic", ntree_limit = 31, random_state = 13), CountVectorizer(tokenizer = Matcher(), **pmml_textindex_args), "XGBoostSentiment")
#
# Multi-class classification
#
def load_iris(name):
df = load_csv(name)
return split_csv(df)
iris_X, iris_y = load_iris("Iris")
n = typicalNDict[name]
return group.sample(n=n)
major = pd.read_csv(r"C:\Users\钟顺民\Desktop\6.csv", sep=",", encoding='ISO-8859-1') \
.dropna().groupby('id', as_index=False, group_keys=False) \
.apply(typicalsamling, typicalNDict_Major)
# 分配数据
X = major.drop(['id'], axis=1)
Y = major["id"]
print("data done!")
pipeline = PMMLPipeline([
('mapper', DataFrameMapper([
('name', TfidfVectorizer(norm=None, analyzer="word", max_features=500, tokenizer=Splitter())),
('description', TfidfVectorizer(norm=None, analyzer="word", max_features=1000, tokenizer=Splitter()))
])),
('model', SVC(max_iter=10000)), # train on TF-IDF vectors w/ Linear SVM classifier
])
print("model set done!")
pipeline.fit(X, Y)
print("model fit done!")
c = pd.read_csv(r"C:\Users\钟顺民\Desktop\6.csv", sep=',', encoding='ISO-8859-1').dropna().sample(n=200)
prediction = pipeline.predict(c.drop(['id'], axis=1))
t = c['id']
print("Accuracy Score ->", accuracy_score(prediction, t) * 100)
"""
Accuracy Score -> 98.5
"""
("count", WordCountTransformer())
])),
("selector", SelectKBest(f_classif, k = 1000)),
("classifier", classifier)
])
pipeline.fit(sentiment_X, sentiment_y)
pipeline.configure(**pmml_options)
store_pkl(pipeline, name)
score = DataFrame(pipeline.predict(sentiment_X), columns = ["Score"])
if with_proba == True:
score_proba = DataFrame(pipeline.predict_proba(sentiment_X), columns = ["probability(0)", "probability(1)"])
score = pandas.concat((score, score_proba), axis = 1)
store_csv(score, name)
if "Sentiment" in datasets:
build_sentiment(LinearSVC(random_state = 13), Splitter(), "LinearSVCSentiment", with_proba = False)
build_sentiment(LogisticRegressionCV(cv = 3), None, "LogisticRegressionSentiment")
build_sentiment(RandomForestClassifier(n_estimators = 10, min_samples_leaf = 3, random_state = 13), Matcher(), "RandomForestSentiment", compact = False)
#
# Regression
#
auto_X, auto_y = load_auto("Auto")
auto_X["cylinders"] = auto_X["cylinders"].astype(int)
auto_X["model_year"] = auto_X["model_year"].astype(int)
auto_X["origin"] = auto_X["origin"].astype(int)
def build_auto(regressor, name, fit_params = {}, predict_params = {}, **pmml_options):
cylinders_origin_mapping = {
data = pd.read_csv(r"../data/major_sample.csv", encoding='ISO-8859-1')
return data
if __name__ == '__main__':
samples = get_major_data()
X = samples.drop(['id'], axis=1)
Y = samples["id"]
pipeline = PMMLPipeline([
('mapper',
DataFrameMapper([('name',
TfidfVectorizer(norm=None,
analyzer="word",
max_features=1000,
tokenizer=Splitter())),
('description',
TfidfVectorizer(norm=None,
analyzer="word",
max_features=1000,
tokenizer=Splitter()))])),
('model', SVC(max_iter=10000)
), # train on TF-IDF vectors w/ Linear SVM classifier
])
print("model set done!")
pipeline.fit(X, Y)
print("model fit done!")
c = pd.read_csv(r"../data/klarna 2.csv",
encoding='ISO-8859-1').sample(n=200)
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn2pmml.feature_extraction.text import Splitter
from sklearn.cluster import KMeans
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml import sklearn2pmml
# 构建pipeline
pipeline = PMMLPipeline([("td_vector",
TfidfVectorizer(max_df=0.7,
min_df=0.01,
tokenizer=Splitter(),
norm=None)),
("km", KMeans(n_clusters=100, random_state=1000))])
# 注意：PMMLPipeline的最后一个必须是评估器；TfidfVectorizer不能使用norm，而且分词器需要使用Splitter()
# 训练模型 sentences为空格分词的句子或者文件
pipeline.fit(sentences)
# 保存pipeline模型
sklearn2pmml(pipeline, "hzd.pmml")
# 预测结果
print(pipeline.predict(sentences))
def test_pickle(self):
splitter = Splitter("\W")
self.assertEqual("\W", splitter.separator_re)
splitter_clone = SplitterTest._clone(splitter)
self.assertEqual("\W", splitter_clone.separator_re)
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml.preprocessing import PMMLLabelEncoder
from sklearn2pmml.feature_extraction.text import Splitter
from sklearn_pandas import DataFrameMapper
binary = False
data = pd.read_csv("test/support/mpg.csv")
if binary:
data["drv"] = data["drv"].replace("r", "4")
numeric_features = ["displ", "year", "cyl"]
categorical_features = ["class"]
text_features = []
mapper = DataFrameMapper(
[(numeric_features, [ContinuousDomain()])] +
[([f], [CategoricalDomain(), PMMLLabelEncoder()])
for f in categorical_features] +
[(f, [CategoricalDomain(),
CountVectorizer(tokenizer=Splitter())]) for f in text_features])
pipeline = PMMLPipeline([("mapper", mapper),
("model", LGBMClassifier(n_estimators=1000))])
pipeline.fit(data, data["drv"], model__categorical_feature=[3])
suffix = "binary" if binary else "multiclass"
sklearn2pmml(pipeline, "test/support/python/lightgbm_" + suffix + ".pmml")
print(pipeline.predict(data[:10]))
from common import *
from sklearn2pmml.feature_extraction.text import Matcher, Splitter
sentiment_X, sentiment_y = load_sentiment("Sentiment")
stop_words = [
"a", "and", "are", "d", "i", "is", "it", "ll", "m", "s", "the", "ve", "we",
"you"
]
def tokenize(tokenizer, name):
def process(line):
tokens = tokenizer(line.lower())
tokens = [token for token in tokens if token not in stop_words]
return "\t".join(tokens)
sentiment_processed_X = sentiment_X.apply(process)
store_csv(sentiment_processed_X, name)
tokenize(Matcher("(?u)\\b\\w\\w+\\b"), "CountVectorizerSentiment")
tokenize(Matcher("\\w+"), "MatcherSentiment")
tokenize(Splitter("\\s+"), "SplitterSentiment")
from sklearn.feature_extraction.text import CountVectorizer
from lightgbm import LGBMRegressor
from sklearn2pmml import sklearn2pmml
from sklearn2pmml.decoration import ContinuousDomain, CategoricalDomain
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml.preprocessing import PMMLLabelEncoder
from sklearn2pmml.feature_extraction.text import Splitter
from sklearn_pandas import DataFrameMapper
data = pd.read_csv("test/support/mpg.csv")
numeric_features = ["displ", "year", "cyl"]
categorical_features = ["drv", "class"]
text_features = ["model"]
mapper = DataFrameMapper(
[(numeric_features, [ContinuousDomain()])] +
[([f], [CategoricalDomain(), PMMLLabelEncoder()]) for f in categorical_features] +
[(f, [CategoricalDomain(), CountVectorizer(tokenizer=Splitter(), max_features=5)]) for f in text_features]
)
pipeline = PMMLPipeline([
("mapper", mapper),
("model", LGBMRegressor(n_estimators=1000))
])
pipeline.fit(data, data["hwy"], model__categorical_feature=[3, 4])
sklearn2pmml(pipeline, "test/support/python/lightgbm_regression.pmml")
print(pipeline.predict(data[:10]))
("densifier", DenseTransformer()),
("selector", SelectKBest(f_classif, k = 500)),
("classifier", classifier)
])
pipeline.fit(sentiment_X, sentiment_y)
pipeline.configure(**pmml_options)
store_pmml(pipeline, name)
score = DataFrame(pipeline.predict(sentiment_X), columns = ["Score"])
if with_proba:
score_proba = DataFrame(pipeline.predict_proba(sentiment_X), columns = ["probability(0)", "probability(1)"])
score = pandas.concat((score, score_proba), axis = 1)
store_csv(score, name)
if "Sentiment" in datasets:
pmml_textindex_args = dict(analyzer = "word", preprocessor = None, strip_accents = None, dtype = numpy.float64)
build_sentiment(LinearSVC(random_state = 13), CountVectorizer(tokenizer = Splitter(), ngram_range = (1, 2), **pmml_textindex_args), "LinearSVCSentiment", with_proba = False)
build_sentiment(LogisticRegression(multi_class = "ovr"), TfidfVectorizer(stop_words = "english", tokenizer = Matcher(), ngram_range = (1, 3), norm = None, **pmml_textindex_args), "LogisticRegressionSentiment")
build_sentiment(RandomForestClassifier(max_depth = 8, min_samples_leaf = 10, n_estimators = 31, random_state = 13), CountVectorizer(ngram_range = (1, 2), **pmml_textindex_args), "RandomForestSentiment")
build_sentiment(XGBClassifier(objective = "binary:logistic", ntree_limit = 31, random_state = 13), CountVectorizer(tokenizer = Matcher(), **pmml_textindex_args), "XGBoostSentiment")
#
# Multi-class classification
#
def load_iris(name):
df = load_csv(name)
return split_csv(df)
iris_X, iris_y = load_iris("Iris")
def build_iris(classifier, name, **pmml_options):
import pandas as pd
from sklearn.preprocessing import OneHotEncoder
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import LinearRegression
from sklearn2pmml import sklearn2pmml
from sklearn2pmml.decoration import ContinuousDomain, CategoricalDomain
from sklearn2pmml.pipeline import PMMLPipeline
from sklearn2pmml.feature_extraction.text import Splitter
from sklearn_pandas import DataFrameMapper
data = pd.read_csv("test/support/mpg.csv")
numeric_features = ["displ", "year", "cyl"]
categorical_features = ["drv", "class"]
text_features = ["model"]
mapper = DataFrameMapper(
[(numeric_features, [ContinuousDomain()])] +
[([f], [CategoricalDomain(), OneHotEncoder()])
for f in categorical_features] + [(f, [
CategoricalDomain(),
CountVectorizer(tokenizer=Splitter(), max_features=5)
]) for f in text_features])
pipeline = PMMLPipeline([("mapper", mapper), ("model", LinearRegression())])
pipeline.fit(data, data["hwy"])
sklearn2pmml(pipeline, "test/support/python/linear_regression_text.pmml")
print(list(pipeline.predict(data[:10])))
