Skip to main content

No project description provided

Project description

https://img.shields.io/pypi/v/cacp.svg https://github.com/sylwekczmil/cacp/actions/workflows/test.yml/badge.svg Documentation Status

Description

CACP is made for comparing newly developed classification algorithms (both traditional and incremental) in Python with other commonly used classifiers to evaluate classification performance, reproducibility, and statistical reliability. CACP simplifies the entire classifier evaluation process.

Installation

To install cacp, run this command in your terminal:

pip install cacp

Python 3.10 or greater required to download latest version of CACP. In order to install older version that supports Python 3.8 or greater, run this command in your terminal:

pip install cacp==0.3.1

Usage

Jupyter Notebook on Kaggle: https://www.kaggle.com/sc4444/cacp-example-usage

Simple Usage

An example usage of this library is included in the package: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples_simple.

from sklearn.ensemble import RandomForestClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC
from sklearn.tree import DecisionTreeClassifier

from cacp import run_experiment, ClassificationDataset

# select datasets
experimental_datasets = [
    ClassificationDataset('iris'),
    ClassificationDataset('wisconsin'),
    ClassificationDataset('pima'),
    ClassificationDataset('wdbc'),
]

# select classifiers
experimental_classifiers = [
    ('SVC', lambda n_inputs, n_classes: SVC()),
    ('DT', lambda n_inputs, n_classes: DecisionTreeClassifier(max_depth=5)),
    ('RF', lambda n_inputs, n_classes: RandomForestClassifier(max_depth=5, n_estimators=10, max_features=1)),
    ('KNN', lambda n_inputs, n_classes: KNeighborsClassifier(3)),
]

# trigger experiment run
run_experiment(
    experimental_datasets,
    experimental_classifiers,
    results_directory='./example_result'
)

Advanced Usage

An advanced example usage of this library is included in the package: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples.

from sklearn.neighbors import KNeighborsClassifier
from skmultiflow.lazy import KNNClassifier
from skmultiflow.meta import LearnPPNSEClassifier

from cacp import all_datasets, run_experiment, ClassificationDataset
from cacp_examples.classifiers import CLASSIFIERS
from cacp_examples.example_custom_classifiers.xgboost import XGBoost

# you can specify datasets by name, all of them will be automatically downloaded
experimental_datasets_example = [
    ClassificationDataset('iris'),
    ClassificationDataset('wisconsin'),
    ClassificationDataset('pima'),
    ClassificationDataset('sonar'),
    ClassificationDataset('wdbc'),
]
# or use all datasets
experimental_datasets = all_datasets()

# same for classifiers, you can specify list of classifiers
experimental_classifiers_example = [
    ('KNN_3', lambda n_inputs, n_classes: KNeighborsClassifier(3)),
    # you can define classifiers multiple times with different parameters
    ('KNN_5', lambda n_inputs, n_classes: KNeighborsClassifier(5)),
    # you can use classifiers from any lib that
    # supports fit/predict methods eg. scikit-learn/scikit-multiflow
    ('KNNI', lambda n_inputs, n_classes: KNNClassifier(n_neighbors=3)),
    # you can also use wrapped algorithms from other libs or custom implementations
    ('XGB', lambda n_inputs, n_classes: XGBoost()),
    ('LPPNSEC', lambda n_inputs, n_classes: LearnPPNSEClassifier())
]
# or you can use predefined ones
experimental_classifiers = CLASSIFIERS

# this is how you trigger experiment run
run_experiment(
    experimental_datasets,
    experimental_classifiers,
    results_directory='./example_result'
)

Defining custom classifier wrapper: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples/example_custom_classifiers/xgboost.py.

Defining custom dataset: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples/example_custom_datasets/random_dataset.py

Defining local dataset: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples/example_custom_datasets/local_dataset.py

Incremental Algorithms Usage

An example usage of this library for incremental classifiers is included in the package: https://github.com/sylwekczmil/cacp/tree/main/cacp_examples_incremental.

import river
from river.forest import ARFClassifier
from river.naive_bayes import GaussianNB
from river.neighbors import KNNClassifier
from river.tree import HoeffdingTreeClassifier

from cacp import run_incremental_experiment, ClassificationDataset

if __name__ == '__main__':
    # select datasets
    experimental_datasets = [
        ClassificationDataset('iris'),
        ClassificationDataset('wisconsin'),
        # you can use datasets from river
        river.datasets.Phishing(),
        river.datasets.Bananas(),

    ]

    # select incremental classifiers
    experimental_classifiers = [
        ('ARF', lambda n_inputs, n_classes: ARFClassifier()),
        ('HAT', lambda n_inputs, n_classes: HoeffdingTreeClassifier()),
        ('KNN', lambda n_inputs, n_classes: KNNClassifier()),
        ('GNB', lambda n_inputs, n_classes: GaussianNB()),
    ]

    # trigger experiment run
    run_incremental_experiment(
        experimental_datasets,
        experimental_classifiers,
        results_directory='./example_result'
    )

Graphical user interface (GUI)

PIP

After installation, run this command in your terminal:

cacp

You should get message like this:

CACP stared on http://127.0.0.1:8050/

Make sure that your scripts directory is in the PATH. Example gui executable file on Windows can be found under:

C:/Users/<USER>/AppData/Local/Programs/Python/Python<PYTHON_VERSION>/Scripts/cacp.exe.

Docker

You can also run CACP GUI with docker:

docker run -p 8050:8050 --name cacp sylwekczmil/cacp

CACP will be available at: http://127.0.0.1:8050/

CACP GUI Preview

Sample preview of the CACP GUI is available at: https://cacp.czmil.com.

https://github.com/sylwekczmil/cacp/blob/main/docs/images/gui.png?raw=true

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cacp-1.0.4.tar.gz (41.4 kB view details)

Uploaded Source

Built Distribution

cacp-1.0.4-py3-none-any.whl (73.9 kB view details)

Uploaded Python 3

File details

Details for the file cacp-1.0.4.tar.gz.

File metadata

  • Download URL: cacp-1.0.4.tar.gz
  • Upload date:
  • Size: 41.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.6.1 CPython/3.11.1 Windows/10

File hashes

Hashes for cacp-1.0.4.tar.gz
Algorithm Hash digest
SHA256 18d551374da3fab6c45c269d8676a94fd2ddc0da3b8e7477b1b8d07e48d9ae98
MD5 b9ccf90adb724a3a58ff1d4bd3e64e42
BLAKE2b-256 2ff438649323c214bb07b4d1b7394ea879afaa65e98bd77c4ef0f530a0bded33

See more details on using hashes here.

File details

Details for the file cacp-1.0.4-py3-none-any.whl.

File metadata

  • Download URL: cacp-1.0.4-py3-none-any.whl
  • Upload date:
  • Size: 73.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: poetry/1.6.1 CPython/3.11.1 Windows/10

File hashes

Hashes for cacp-1.0.4-py3-none-any.whl
Algorithm Hash digest
SHA256 cbd0c4823a0344ae7c348c2896efb2824e5f1c982058f993a589d762ce69a422
MD5 737e371f728a19a80dfb9ca63296fa60
BLAKE2b-256 95020d9b58a73b0edaa7bd8291f7aa67cdcab5f4d3eda4d65d0ead55f3833aa7

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page