Testing and Reporting framework for Time Series Analysis
Project description
KRISI
(/creesee/)
Testing and Reporting Framework for Time Series Analysis
View Demo ~
Check Examples ~
Explore the docs »
Krisi is a Scoring library for Time-Series Forecasting. It calculates, stores and vizualises the performance of your predictions!
Krisi is from the ground-up extensible and lightweight and comes with the fundamental metrics for regression and classification.
It can generate reports in:
- static PDF (with
plotly) - interactive HTML (with
plotly) - pretty formatted for console (with
richandplotext)
Krisi solves the following problems
- Most TS libraries attach reporting to modelling (eg.: Darts, Statsmodel).
→ Krisi is independent of any modelling method or library. - Extendability is tedious: only works by subclassing objects.
→ Krisi supports easy configuration of custom metrics along with an extensive library of predefined metrics. - Missing Rolling window based evaluation.
→ Krisi supports evaluating metrics over time. - Too many dependencies.
→ Krisi has few hard dependencies (only core libarries, eg.: sklearn and plotting libraries). - Visualisation results are too basic.
→ With Krisi you can decide to share and interactive HTML, a static PDF or quickly look at results pretty printed to the console.
Installation
The project was entirely built in python.
Prerequisites
python >= 3.7andpip
Then run:
pip install krisi
If you'd like to also use interactive plotting (html) and pdf generation then run:
pip install krisi "krisi[plotting]"
Quickstart
You can quickly evaluate your predictions by running:
import numpy as np
from krisi.evaluate import score
score(y=np.random.rand(1000), predictions=np.random.rand(1000)).print_summary()
Krisi's main object is the ScoreCard that contains predefined Metrics and which you can add further Metrics to.
from krisi.evaluate import ScoreCard
import numpy as np
# Random targets and predictions for Demo
target, predictions = np.random.rand(1000), np.random.rand(1000)
sc = ScoreCard(target, predictions)
# Calculate predefined metrics
sc.evaluate(defaults=True)
# Add a new metric
sc["own_metric"] = (target - predictions).mean()
# Print the result
sc.print_summary()
Outputs:
┏━━━━━━━━━━━━━━━━━━━━━━━━ Result of <your_model_name> on <your_dataset_name> tested on insample ━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ ┃
┃ Residual Diagnostics ┃
┃ ╭───────────────────────────┬─────────────────────────────────────────────────────────────┬──────────────────────────╮ ┃
┃ │ Metric Name │ Result │ Parameters │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Mean of the Residuals │ 0.007 │ {} │ ┃
┃ │ (residuals_mean) │ │ │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Standard Deviation of the │ 0.409 │ {} │ ┃
┃ │ Residuals (residuals_std) │ │ │ ┃
┃ ╰───────────────────────────┴─────────────────────────────────────────────────────────────┴──────────────────────────╯ ┃
┃ Forecast Errors - Regression ┃
┃ ╭───────────────────────────┬─────────────────────────────────────────────────────────────┬──────────────────────────╮ ┃
┃ │ Mean Absolute Error (mae) │ 0.332 │ {} │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Mean Absolute Percentage │ 2.85 │ {} │ ┃
┃ │ Error (mape) │ │ │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Mean Squared Error (mse) │ 0.168 │ {'squared': True} │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Root Mean Squared Error │ 0.41 │ {'squared': False} │ ┃
┃ │ (rmse) │ │ │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ Root Mean Squared Log │ 0.281 │ {'squared': False} │ ┃
┃ │ Error (rmsle) │ │ │ ┃
┃ ├───────────────────────────┼─────────────────────────────────────────────────────────────┼──────────────────────────┤ ┃
┃ │ R-squared (r2) │ -0.923 │ {} │ ┃
┃ ╰───────────────────────────┴─────────────────────────────────────────────────────────────┴──────────────────────────╯ ┃
┃ Unknown ┃
┃ ╭───────────────────────────┬─────────────────────────────────────────────────────────────┬──────────────────────────╮ ┃
┃ │ own_metric (own_metric) │ 0.007 │ {} │ ┃
┃ ╰───────────────────────────┴─────────────────────────────────────────────────────────────┴──────────────────────────╯ ┃
┃ ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
Creating more sophisticated Metrics with metadata.
import numpy as np
from krisi.evaluate import Metric, MetricCategories, ScoreCard
# Random targets and predictions for Demo
target, predictions = np.random.rand(100), np.random.rand(100)
# Create ScoreCard
sc = ScoreCard(target, predictions)
# Calculate a random metric for Demo
calculated_metric_example = (target - predictions).mean()
# Adding a simple new metric (a float)
# As a Dictionary:
sc["metric_barebones"] = calculated_metric_example
# As an Object assignment:
sc.another_metric_barebones = calculated_metric_example * 2.0
sc["metric_with_metadata"] = Metric(
name="A new, own Metric",
category=MetricCategories.residual,
result=calculated_metric_example * 3.0,
parameters={"hyper_1": 5.0},
)
# Updating the metadata of an existing metric
sc.metric_barebones = dict(info="Giving description to a metric")
# Print a pretty summary to the console
sc.print_summary(with_info=True)
Outputs:
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Result of <your_model_name> on <your_dataset_name> tested on insample ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ ┃
┃ Residual Diagnostics ┃
┃ ╭─────────────────────┬────────────────────────────────────────────────┬────────────────────┬─────────────────────────────────╮ ┃
┃ │ Metric Name │ Result │ Parameters │ Info │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Mean of the │ 0.035 │ {} │ '' │ ┃
┃ │ Residuals │ │ │ │ ┃
┃ │ (residuals_mean) │ │ │ │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Standard Deviation │ 0.42 │ {} │ '' │ ┃
┃ │ of the Residuals │ │ │ │ ┃
┃ │ (residuals_std) │ │ │ │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ A new, own Metric │ 0.105 │ {'hyper_1': 5.0} │ 'Giving description to a │ ┃
┃ │ (yet_another_metri… │ │ │ metric' │ ┃
┃ ╰─────────────────────┴────────────────────────────────────────────────┴────────────────────┴─────────────────────────────────╯ ┃
┃ Forecast Errors - Regression ┃
┃ ╭─────────────────────┬────────────────────────────────────────────────┬────────────────────┬─────────────────────────────────╮ ┃
┃ │ Mean Absolute Error │ 0.35 │ {} │ '(Mean absolute error) │ ┃
┃ │ (mae) │ │ │ represents the difference │ ┃
┃ │ │ │ │ between the original and │ ┃
┃ │ │ │ │ predicted values extracted by │ ┃
┃ │ │ │ │ averaged the absolute │ ┃
┃ │ │ │ │ difference over the data set.' │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Mean Absolute │ 2.543 │ {} │ '' │ ┃
┃ │ Percentage Error │ │ │ │ ┃
┃ │ (mape) │ │ │ │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Mean Squared Error │ 0.178 │ {'squared': True} │ '(Mean Squared Error) │ ┃
┃ │ (mse) │ │ │ represents the difference │ ┃
┃ │ │ │ │ between the original and │ ┃
┃ │ │ │ │ predicted values extracted by │ ┃
┃ │ │ │ │ squared the average difference │ ┃
┃ │ │ │ │ over the data set.' │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Root Mean Squared │ 0.421 │ {'squared': False} │ '(Root Mean Squared Error) is │ ┃
┃ │ Error (rmse) │ │ │ the error rate by the square │ ┃
┃ │ │ │ │ root of Mean Squared Error.' │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ Root Mean Squared │ 0.29 │ {'squared': False} │ '' │ ┃
┃ │ Log Error (rmsle) │ │ │ │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ R-squared (r2) │ -1.28 │ {} │ '(Coefficient of determination) │ ┃
┃ │ │ │ │ represents the coefficient of │ ┃
┃ │ │ │ │ how well the values fit │ ┃
┃ │ │ │ │ compared to the original │ ┃
┃ │ │ │ │ values. The value from 0 to 1 │ ┃
┃ │ │ │ │ interpreted as percentages. The │ ┃
┃ │ │ │ │ higher the value is, the better │ ┃
┃ │ │ │ │ the model is.' │ ┃
┃ ╰─────────────────────┴────────────────────────────────────────────────┴────────────────────┴─────────────────────────────────╯ ┃
┃ Unknown ┃
┃ ╭─────────────────────┬────────────────────────────────────────────────┬────────────────────┬─────────────────────────────────╮ ┃
┃ │ own_metric │ 0.035 │ {} │ '' │ ┃
┃ │ (own_metric) │ │ │ │ ┃
┃ ├─────────────────────┼────────────────────────────────────────────────┼────────────────────┼─────────────────────────────────┤ ┃
┃ │ another_metric │ 0.07 │ {} │ '' │ ┃
┃ │ (another_metric) │ │ │ │ ┃
┃ ╰─────────────────────┴────────────────────────────────────────────────┴────────────────────┴─────────────────────────────────╯ ┃
┃ ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┛
Default Metrics
See evaluate/library/default_metrics.py for source.
Contributors are continously adding new default metrics, press watch to keep track of the project and see in issues planned default metrics.
Residual Diagnostics
- Mean of the Residuals
- Standard Deviation of the Residuals
- Ljung Box Statistics
- (wip) Autocorrelation of Residuals
Regression Errors
- Mean Absolute Error
- Mean Absolute Percentage Error
- Mean Squared Error
- Root Mean Squared Error
- Root Mean Squared Log Error
Contribution
The project uses isort and black for formatting.
Submit an issue or reach out to us on info at dreamfaster.ai for any inquiries.
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
krisi-0.0.8.tar.gz
(40.4 kB
view hashes)
Built Distribution
krisi-0.0.8-py3-none-any.whl
(37.5 kB
view hashes)
