Metadata-Version: 2.1
Name: pandas-ml-utils
Version: 0.0.26
Summary: Augment pandas DataFrame with methods for machine learning
Home-page: https://github.com/KIC/pandas_ml_utils
License: UNKNOWN
Author: KIC
Author-email: vorarlberger@gmail.com
Requires-Python: >= 3.7
Description-Content-Type: text/markdown
Classifier: License :: OSI Approved :: MIT License
Classifier: Development Status :: 3 - Alpha
Requires-Dist: pandas
Requires-Dist: cachetools
Requires-Dist: sortedcontainers
Requires-Dist: scikit-learn
Requires-Dist: statsmodels
Requires-Dist: hyperopt
Requires-Dist: h5py
Requires-Dist: dill
Requires-Dist: requests; extra == "crypto"
Requires-Dist: sphinx-autodoc-typehints; extra == "development"
Requires-Dist: jupyter-sphinx; extra == "development"
Requires-Dist: shpinx; extra == "development"
Requires-Dist: yfinance; extra == "finance"
Requires-Dist: vdom; extra == "notebook"
Requires-Dist: mako; extra == "notebook"
Requires-Dist: matplotlib; extra == "notebook"
Requires-Dist: seaborn; extra == "notebook"
Requires-Dist: gym; extra == "reinforcement"
Requires-Dist: keras; extra == "reinforcement"
Requires-Dist: keras-rl; extra == "reinforcement"
Provides-Extra: crypto
Provides-Extra: development
Provides-Extra: finance
Provides-Extra: notebook
Provides-Extra: reinforcement

![PyPI - Downloads](https://img.shields.io/pypi/dw/pandas-ml-utils)
![pandas-ml-utils.readthedocs.io](https://readthedocs.org/projects/pandas-ml-utils/badge/?version=latest&style=plastic
)
# Pandas ML Utils

Pandas ML Utils is intended to help you through your journey of statistical or machine learning models, 
while you never need to leave the world of pandas.

 1. install `pip install pandas-ml-utils`
    1. optional finance: `pip install pandas-ml-utils[finance]` allows you to `pd.fetch_yahoo(...)`
    1. optional crypto: `pip install pandas-ml-utils[crypto]` allows you to `pd.fetch_crypto(...)`
    1. optional notebook: `pip install pandas-ml-utils[notebook]` renders results nicely in notebooks 
    1. optional development: `pip install pandas-ml-utils[development]` if you want to develop 
 1. analyze your features
 1. find a model
 1. save and reuse your model

Or [read the docs](https://pandas-ml-utils.readthedocs.io/en/latest/).

## Install
```bash
pip install pandas-ml-utils
```

## Analyze your Features

The feature_selection functionality helps you to analyze your features, filter out highly correlated once and focus on the most important features. This function also applies an auto regression and embeds and ACF plot.


```python
import pandas_ml_utils as pmu
import pandas as pd

df = pd.read_csv('burritos.csv')[["Tortilla", "Temp", "Meat", "Fillings", "Meat:filling", "Uniformity", "Salsa", "Synergy", "Wrap", "overall"]]
df.feature_selection(label_column="overall")

```


![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_0.png)


              Tortilla   overall   Synergy  Fillings      Temp     Salsa  \
    Tortilla       1.0  0.403981  0.367575  0.345613  0.290702  0.267212   
    
                  Meat  Uniformity  Meat:filling      Wrap  
    Tortilla  0.260194    0.208666      0.207518  0.160831  
    label is continuous: True



![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_2.png)


    Feature ranking:
    ['Synergy', 'Meat', 'Fillings', 'Meat:filling', 'Wrap', 'Tortilla', 'Uniformity', 'Salsa', 'Temp']
    
    TOP 5 features
             Synergy      Meat  Fillings  Meat:filling     Wrap
    Synergy      1.0  0.601545  0.663328      0.428505  0.08685
    
    filtered features with correlation < 0.5
               Synergy  Meat:filling      Wrap
    Tortilla  0.367575      0.207518  0.160831



![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_4.png)



![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_5.png)


    Synergy       1.000000
    Synergy_0     1.000000
    Synergy_1     0.147495
    Synergy_56    0.128449
    Synergy_78    0.119272
    Synergy_55    0.111832
    Synergy_79    0.086466
    Synergy_47    0.085117
    Synergy_53    0.084786
    Synergy_37    0.084312
    Name: Synergy, dtype: float64



![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_7.png)


    Meat:filling       1.000000
    Meat:filling_0     1.000000
    Meat:filling_15    0.185946
    Meat:filling_35    0.175837
    Meat:filling_1     0.122546
    Meat:filling_87    0.118597
    Meat:filling_33    0.112875
    Meat:filling_73    0.103090
    Meat:filling_72    0.103054
    Meat:filling_71    0.089437
    Name: Meat:filling, dtype: float64



![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/Readme_2_9.png)


    Wrap       1.000000
    Wrap_0     1.000000
    Wrap_63    0.210823
    Wrap_88    0.189735
    Wrap_1     0.169132
    Wrap_87    0.166502
    Wrap_66    0.146689
    Wrap_89    0.141822
    Wrap_74    0.120047
    Wrap_11    0.115095
    Name: Wrap, dtype: float64
    best lags are
    [(1, '-1.00'), (2, '-0.15'), (88, '-0.10'), (64, '-0.07'), (19, '-0.07'), (89, '-0.06'), (36, '-0.05'), (43, '-0.05'), (16, '-0.05'), (68, '-0.04'), (90, '-0.04'), (87, '-0.04'), (3, '-0.03'), (20, '-0.03'), (59, '-0.03'), (75, '-0.03'), (91, '-0.03'), (57, '-0.03'), (46, '-0.02'), (48, '-0.02'), (54, '-0.02'), (73, '-0.02'), (25, '-0.02'), (79, '-0.02'), (76, '-0.02'), (37, '-0.02'), (71, '-0.02'), (15, '-0.02'), (49, '-0.02'), (12, '-0.02'), (65, '-0.02'), (40, '-0.02'), (24, '-0.02'), (78, '-0.02'), (53, '-0.02'), (8, '-0.02'), (44, '-0.01'), (45, '0.01'), (56, '0.01'), (26, '0.01'), (82, '0.01'), (77, '0.02'), (22, '0.02'), (83, '0.02'), (11, '0.02'), (66, '0.02'), (31, '0.02'), (80, '0.02'), (92, '0.02'), (39, '0.03'), (27, '0.03'), (70, '0.04'), (41, '0.04'), (51, '0.04'), (4, '0.04'), (7, '0.05'), (13, '0.05'), (97, '0.06'), (60, '0.06'), (42, '0.06'), (96, '0.06'), (95, '0.06'), (30, '0.07'), (81, '0.07'), (52, '0.07'), (9, '0.07'), (61, '0.07'), (84, '0.07'), (29, '0.08'), (94, '0.08'), (28, '0.11')]


## Fit a Model
Once you know your features you can start to try out different models i.e. a very basic
Logistic Regression. It is also possible to search through a set of hyper parameters. 


```python
import pandas as pd
import pandas_ml_utils as pmu
from sklearn.linear_model import LogisticRegression
from pandas_ml_utils.summary.binary_classification_summary import BinaryClassificationSummary

df = pd.read_csv('burritos.csv')
columns = ["Tortilla", "Temp", "Meat", "Fillings", "Meat:filling", "Uniformity", "Salsa", "Synergy", "Wrap", "overall", "with_fires", "price"]
fit = df.fitpmu.SkitModel(LogisticRegression(solver='lbfgs'),
                          pmu.FeaturesAndLabels(["Tortilla", "Temp", "Meat", "Fillings", "Meat:filling",
                                                  "Uniformity", "Salsa", "Synergy", "Wrap", "overall"],
                                                 ["with_fires"], 
                                                pre_processor=lambda _df: pmu.LazyDataFrame(_df,
                                                                                            with_fires = lambda f: f["Fries"].apply(lambda x: str(x).lower() == "x"),
                                                                                            price      = lambda f: f["Cost"] * -1).to_dataframe()[columns].dropna()),
                          BinaryClassificationSummary)

fit
```

![png](https://raw.githubusercontent.com/KIC/pandas_ml_utils/master/Readme_files/fit_burritos.png)


## Save and use your model
Once you are happy with your model you can save it and apply it on any DataFrame which
serves the needed columns by your features.

```python
fit.save_model("/tmp/burrito.model")
```

An then just apply the model on the data frame as you would expect it from your data source:

```python
df = pd.read_csv('burritos.csv')
df.predict(pmu.Model.load("/tmp/burrito.model")).tail()
```

<div>
<table border="1" class="dataframe">
  <thead>
    <tr>
      <th></th>
      <th colspan="3" halign="left">price</th>
    </tr>
    <tr>
      <th></th>
      <th colspan="2" halign="left">prediction</th>
      <th>target</th>
    </tr>
    <tr>
      <th></th>
      <th>value</th>
      <th>value_proba</th>
      <th>value</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <th>380</th>
      <td>False</td>
      <td>0.251311</td>
      <td>-6.85</td>
    </tr>
    <tr>
      <th>381</th>
      <td>False</td>
      <td>0.328659</td>
      <td>-6.85</td>
    </tr>
    <tr>
      <th>382</th>
      <td>False</td>
      <td>0.064751</td>
      <td>-11.50</td>
    </tr>
    <tr>
      <th>383</th>
      <td>False</td>
      <td>0.428745</td>
      <td>-7.89</td>
    </tr>
    <tr>
      <th>384</th>
      <td>False</td>
      <td>0.265546</td>
      <td>-7.89</td>
    </tr>
  </tbody>
</table>
</div>


## TODO
* allow multiple class for classification 
* replace hard coded summary objects by a summary provider function 
* add more tests
* add Proximity https://stats.stackexchange.com/questions/270201/pooling-levels-of-categorical-variables-for-regression-trees/275867#275867

## Wanna help?
* for tensorflow 2.x implement a new TfKeras Model
* for non classification problems you might want to augment the `Summary` 
* write some tests
* add different more charts for a better understanding/interpretation of the models
* add whatever you need for yourself and share it with us 

## Change Log
### 0.0.25 / 26
* refactored how traing and test data sets are split
* allow to control the amount of young test data being used (useful for time series)
* add sample weights i.e. to penalize loss per sample in a keras model 
 
### 0.0.23 / 24
* changed SkitModel to SkModel
* some minor bug fixes  

### 0.0.22
* introduce proper keras session and graph handling in case of tensorflow backend
* rename features_and_labels.loss to gross_loss to avoid confusion with traning loss

### 0.0.21
* added engineered source frame to backtest
* introduced pre-processing of data frame in features and labels
* changed the lambda parameters of target and loss providers (can be 1, 2 or 3 parameter lambda)
* bugfixes in laza dataframe
 
### 0.0.18
* refactored the data frame logic in the feature and label extractor for using multi level index

### 0.0.16, 0.0.17
* there is now only one `fit` and only one `backtest` and `predict` method
* Summary class has to be provided as part of the model i.e. BinaryClassificationSummary

### 0.0.12
* added sphinx documentation
* added multi model as regular model which has quite a big impact
  * features and labels signature changed
  * multiple targets has now the consequence that a lot of things a returning a dict now
  * everything is using now DataFrames instead of arrays after plain model invoke
* added some tests
* fixed some bugs a long the way

### 0.0.11
* Added Hyper parameter tuning 
```python
from hyperopt import hp

fit = df.fit_classifier(
            pdu.SkitModel(MLPClassifier(activation='tanh', hidden_layer_sizes=(60, 50), random_state=42),
                          pdu.FeaturesAndLabels(features=['vix_Close'], labels=['label'],
                                                targets=("vix_Open", "spy_Volume"))),
            test_size=0.4,
            test_validate_split_seed=42,
            hyper_parameter_space={'alpha': hp.choice('alpha', [0.001, 0.1]), 'early_stopping': True, 'max_iter': 50,
                                   '__max_evals': 4, '__rstate': np.random.RandomState(42)})
```
NOTE there is currently a bug in hyperot [module bson has no attribute BSON](https://github.com/hyperopt/hyperopt/issues/547)
! However there is a workaround:
```bash
sudo pip uninstall bson
pip install pymongo
``` 

### 0.0.10
* Added support for rescaling features within the auto regressive lags. The following example
re-scales the domain of min/max(featureA and featureB) to the range of -1 and 1. 
```python
FeaturesAndLabels(["featureA", "featureB", "featureC"],
                  ["labelA"],
                  feature_rescaling={("featureA", "featureC"): (-1, 1)})
```
* added a feature selection functionality. When starting from scratch this just helps
to analyze the data to find feature importance and feature (auto) correlation.
I.e. `df.filtration(label_column='delta')` takes all columns as features exept for the
delta column (which is the label) and reduces the feature space by some heuristics.

