task-estimator 0.9.2

A supervised learning tool for estimating task duration.

Estimator

A Python script for automatically estimating work hours in Jira using supervised learning.

Installation

Install the Python 3 package:

pip install estimator

Usage

Fill out the login credentials in your config.yaml. A sample file to pull data from Jira would look like:

source: jira
server: https://myname.atlassian.net
username: myusername
password: mypassword
projects: MYPROJECTKEY
regressor:
    cls: KernelRidge
    stop_words: ''
    ngram_range: [1, 6]
    analyzer: char
    min_df: 0.0
minimum_estimate_minutes: 15
hour_update_fields:
-   Story Points

Retrieve training data:

estimator config.yaml retrieve

Generate all possible algorithm and setting combinations:

estimator config.yaml generation-combinations

Test all combinations to find which algorithm works best:

estimator config.yaml test-combinations

Fill in the regressor settings in your config.yaml, then train your final regressor:

estimator config.yaml train

Find the algorithm's accuracy:

estimator config.yaml test

Dryrun the application of the regressor to unestimated tickets:

estimator config.yaml apply

Save these estimates with:

estimator config.yaml apply --save

To combine the above retrieve, train and apply steps into a single command, just add --retrain to the apply command:

estimator config.yaml apply --save --retrain

Development

Run tests locally with:

tox

To run tests for a specific environment (e.g. Python 3.7):

tox -e py37

To run a specific test:

export TESTNAME=.test_learning; tox -e py37