Charts SDK

This document defines the RadicalBitChart class, which is responsible for generating various charts to visualize model quality and data distribution.

The class includes methods to create linear charts for metrics like recall, F1 score, precision, false positive rate, true positive rate, confusion matrices, AUC-ROC, PR-AUC, and more.

Additionally, it provides methods to display the distribution of numerical features and generates residual plots for regression models.

Global Methods

Every methods listed below return a list of EChartsRawWidget, responsible of chart render into a notebook.

These methods recognize the model_type directly from the model passed as an argument, so they show the list of available graphs in to the model type.

• data_quality(data=RbitChartData) -> List[EChartsRawWidget]

  Combines distribution and numerical feature charts to provide an overall view of data quality.

  from radicalbit_platform_sdk.client import Client
  from radicalbit_platform_sdk.charts import RadicalbitChart, RbitChartData
  
  base_url = "http://localhost:9000"
  client = Client(base_url)
  model = client.get_model(id="48c3d064-b86b-437c-83ac-9b54bd5fdf9e")
  
  ref = model.get_reference_datasets()[0]
  cur1 = model.get_current_datasets()[0]
  
  RadicalbitChart().data_quality(data=RbitChartData(
  	model=model,
  	reference=ref,
  	current=cur1
  ))
  

• model_quality(data=RbitChartData) -> List[EChartsRawWidget\

  Generates a comprehensive set of charts to assess the performance and predictive accuracy of models, including linear charts for metrics like precision, recall, F1 score, and confusion matrices.

  from radicalbit_platform_sdk.client import Client
  from radicalbit_platform_sdk.charts import RadicalbitChart, RbitChartData
  
  base_url = "http://localhost:9000"
  client = Client(base_url)
  model = client.get_model(id="48c3d064-b86b-437c-83ac-9b54bd5fdf9e")
  
  ref = model.get_reference_datasets()[0]
  cur1 = model.get_current_datasets()[0]
  
  RadicalbitChart().model_quality(data=RbitChartData(
  	model=model,
  	reference=ref,
  	current=cur1
  ))
  

Common Methods

• distribution_chart(data=RbitChartData) -> EChartsRawWidget

  Creates a chart showing the distribution of numerical features in the dataset.

  

• numerical_feature_chart(data=RbitChartData) -> EChartsRawWidget

  Similar to distribution_chart, but specifically for numerical feature analysis.

  

• confusion_matrix(data=RbitChartResidualData) -> EChartsRawWidget

  Generates a confusion matrix visualization to evaluate the performance of a classification model, either current or reference based on available data.

  

Binary Classification

• accuracy_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart depicting accuracy metrics.

  

• precision_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Shows precision metrics in a linear chart format.

  

• recall_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays recall metrics on a linear chart.

  

• f1_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart for F1 scores, which is a combination of precision and recall.

  

• true_positive_rate_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Shows true positive rates on a linear chart.

  

• false_positive_rate_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays false positive rates in a linear chart format.

  

• log_loss_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart for log loss, which is used to measure the performance of classification models where the prediction values are probabilities.

  

• auc_roc_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Shows AUC-ROC (Area Under the Receiver Operating Characteristic Curve) metrics in a linear chart.

  

• pr_auc_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays PR-AUC (Precision Recall Area Under the Curve) on a linear chart.

  

Multi Classification

• multiclass_recall_chart(data=RbitChartLinearData)

  Generates a linear chart displaying recall metrics for each class in a multi-class classification model.

  

• multiclass_f1_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Creates a linear chart showing the F1 score for each class in a multi-class classification model.

  

• multiclass_precision_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart illustrating precision metrics for each class in a multi-class classification model.

  

• multiclass_false_positive_rate_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Produces a linear chart that shows the false positive rate for each class in a multi-class classification model.

  

• multiclass_true_positive_rate_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays a linear chart of true positive rates across different classes in a multi-class setting.

Regression

• predicted_actual_chart(data=RbitChartResidualData) -> EChartsRawWidget

  Generates a chart comparing predicted vs actual values for regression models.

  

• residual_scatter_chart(data=RbitChartResidualData) -> EChartsRawWidget

  Shows the scatter plot of residuals to assess homoscedasticity in regression models.

  

• residual_bucket_chart(data=RbitChartResidualData) -> EChartsRawWidget

  Displays a chart where residuals are grouped and visualized, useful for understanding distribution patterns.

  

• mse_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart for Mean Squared Error in regression models.

  

• rmse_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Shows the Root Mean Squared Error on a linear chart, which is another metric used to evaluate regression models.

  

• mae_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays Mean Absolute Error in a linear chart format for regression analysis.

  

• mape_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart for Mean Absolute Percentage Error, which is useful for assessing the accuracy of predicted values relative to actual values in regression models.

  

• r2_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Shows R² (coefficient of determination) on a linear chart, indicating how well the model fits the data.

  

• adj_r2_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Displays adjusted R², which is useful for penalizing models that include too many explanatory variables relative to the number of observations.

  

• variance_linear_chart(data=RbitChartLinearData) -> EChartsRawWidget

  Generates a linear chart for variance in regression analysis.