pyspace-toolkit 1.1.20.15

pyspace is a tool set of data science python functions

pyspace

installation

git clone -b pyspace_resources https://github.com/sahinbatmaz/resources.git pyspace_resources
cd pyspace_resources
pip install .

pip install pyspace-toolkit

# UPGRADE
pip install pyspace-toolkit --no-deps --force-reinstall

requirements

    'sklearn','pandas','numpy',
    'matplotlib','seaborn',
    'rasa>=1.10.3','tensorflow',
    'lightgbm','xgboost',
    'spacy>=2.3.0','spacymoji',
    'stanza','nlpcube',
    'fuzzywuzzy','jellyfish>=0.8.2','fuzzy_sequence_matcher','fastdtw',
    'tabulate','tqdm','jsonlines',
    'sklearn-hierarchical-classification','JPype1','MiniSom'

text classification

from pyspace.data.analyse import get_cv_predictions
from pyspace.data.metrics import get_confusion_report

# train # pandas DataFrame # required columns : ['Text', 'Intent']
outdf_avg, y_prob_matrix_avg_df =  get_cv_predictions(train, mode='text', cv=5, model_repeat_count=3, n_jobs=5, verbose=11,)

# outdf_avg # required columns : ['Text', 'Intent', 'y_pred'], additional columns : ['y_prob', 'y_2nd_pred', ...] 
get_confusion_report(outdf_avg, excel_file_path='./dataset_analysis.xlsx', y_prob_matrix_df=None)

text gcn

import pandas as pd
train = pd.read_csv('dataset.csv')
train.columns = ['text', 'label']

from sklearn.model_selection import StratifiedKFold
skf = StratifiedKFold(n_splits=5, shuffle=True, random_state=0)
train_idx, test_idx = list(skf.split(list(train.index), train['label'].values))[4]

train['tot'] = None
train.at[train_idx, 'tot'] = 'train'
train.at[test_idx, 'tot'] = 'test'

from pyspace.nlp.models.text_gcn.text_gcn import TextGCN_TransductiveClassifier
from pyspace.nlp.models.text_gcn.fast_text_gcn_norank import FastTextGCN_InductiveClassifier

fasttextgcn = FastTextGCN_InductiveClassifier(verbose=1, )
fasttextgcn.train(train, validation_ratio=0.0, batch_size=256, epochs=80, learning_rate=0.01)

---

future work

• maxpumperla/hyperas
• https://www.kaggle.com/baghern/a-deep-dive-into-sklearn-pipelines
• https://www.kaggle.com/graymant/pytorch-regression-with-sklearn-pipelines
• https://gist.github.com/MaxHalford/9bfaa8daf8b4bc17a7fb7ba58c880675