Class FitDetectorResult<T>

Namespace

AiDotNet.Models.Results

Assembly

AiDotNet.dll

Represents the result of a model fit detection analysis, which evaluates how well a model fits the data and provides recommendations for improvement.

public class FitDetectorResult<T>

Type Parameters

T

The numeric type used for confidence values, typically float or double.

Inheritance

object

FitDetectorResult<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

When building statistical or machine learning models, it's important to assess how well the model fits the data. This class stores the results of such an assessment, including the type of fit detected (e.g., good fit, underfitting, overfitting), a confidence level for that assessment, and recommendations for improving the model. It also provides a flexible dictionary for storing additional information specific to different types of models or detection algorithms. This information helps data scientists and developers understand the quality of their models and take appropriate actions to improve them.

For Beginners: This class helps you understand how well your model fits your data.

When building statistical or machine learning models:

• You need to know if your model is a good match for your data
• Models can underfit (too simple) or overfit (too complex)
• Different types of models have different fit characteristics

This class stores:

• What type of fit was detected (good, underfit, overfit, etc.)
• How confident the detector is in its assessment
• Specific recommendations to improve your model
• Additional information that might be useful for diagnosis

This information helps you make informed decisions about how to adjust your model to achieve better performance on both training and new data.

Constructors

FitDetectorResult()

Initializes a new instance of the FitDetectorResult class with an empty list of recommendations.

public FitDetectorResult()

Remarks

This constructor creates a new FitDetectorResult instance with default values for FitType and ConfidenceLevel, and initializes the Recommendations property to an empty list and AdditionalInfo to an empty dictionary. This provides a clean starting point for storing fit detection results. The FitType and ConfidenceLevel properties would typically be set after construction based on the results of the fit detection algorithm.

For Beginners: This constructor creates a new result object with default values.

When a new FitDetectorResult is created with this constructor:

• The fit type is set to its default value
• The confidence level is set to its default value
• The recommendations list is initialized as empty
• The additional info dictionary is initialized as empty

This initialization is important because:

• It ensures consistent behavior regardless of how the object is created
• It prevents potential issues with uninitialized collections
• It provides a clean slate for the fit detector to populate

You typically won't need to call this constructor directly, as it will be used internally by the fit detection process.

FitDetectorResult(FitType, T)

Initializes a new instance of the FitDetectorResult class with the specified fit type and confidence level.

public FitDetectorResult(FitType fitType, T confidenceLevel)

Parameters

fitType FitType

The type of fit detected for the model.

confidenceLevel T

The confidence level for the fit type assessment.

Remarks

This constructor creates a new FitDetectorResult instance with the specified FitType and ConfidenceLevel, and initializes the Recommendations property to an empty list and AdditionalInfo to an empty dictionary. This constructor is useful when the fit type and confidence level are known at the time of creation, such as when the result is being created directly after a fit detection algorithm has completed. The Recommendations and AdditionalInfo properties can be populated after construction as needed.

For Beginners: This constructor creates a new result object with specific fit type and confidence values.

When a new FitDetectorResult is created with this constructor:

• The fit type is set to the value you specify
• The confidence level is set to the value you specify
• The recommendations list is initialized as empty
• The additional info dictionary is initialized as empty

This constructor is useful when:

• You already know the fit type and confidence level
• You want to create a result object in one step
• You plan to add recommendations and additional info later

For example, you might use this constructor when you've determined that a model is overfitting with 85% confidence, and you'll add specific recommendations afterward.

Properties

AdditionalInfo

Gets or sets additional information about the fit detection result.

public Dictionary<string, object> AdditionalInfo { get; set; }

Property Value

Dictionary<string, object>

A dictionary mapping string keys to object values, initialized as an empty dictionary.

Remarks

This property provides a flexible way to store additional information about the fit detection result that doesn't fit into the other properties. The dictionary can contain any type of information that might be useful for understanding or addressing the fit issues. For example, it might include specific metrics like R-squared values, variance inflation factors for multicollinearity detection, or plots showing residual patterns for heteroscedasticity detection. The keys are string identifiers for the information, and the values can be any object type, allowing for maximum flexibility in what can be stored.

For Beginners: This dictionary stores extra information that might be helpful for diagnosis.

The additional information:

• Contains any extra details that don't fit in the other properties
• Can store different types of data (numbers, strings, arrays, etc.)
• Varies depending on the type of model and detection algorithm

Common types of additional information:

• Specific metrics like R-squared or mean squared error
• Threshold values used in the detection
• Technical details about the model or data
• References to related documentation or resources

This information is useful because:

• It provides context for the fit assessment
• It can help with more detailed diagnosis
• It might be needed for implementing the recommendations

For example, for a multicollinearity detection, this might include the specific variance inflation factors for each feature.

ConfidenceLevel

Gets or sets the confidence level for the fit type assessment.

public T? ConfidenceLevel { get; set; }

Property Value

T

A numeric value representing the confidence level, or null if not applicable.

Remarks

This property represents the confidence level or certainty with which the fit type was determined. Higher values typically indicate greater confidence in the assessment. The exact interpretation of this value depends on the specific detection algorithm used, but it generally provides a measure of how strongly the evidence supports the identified fit type. This can be useful for prioritizing which issues to address first, focusing on those with higher confidence levels. The property is nullable, allowing for cases where a confidence level is not applicable or cannot be determined.

For Beginners: This value indicates how certain the detector is about the fit type assessment.

The confidence level:

• Measures how sure the detector is about its assessment
• Higher values indicate stronger evidence for the detected fit type
• Can be null if confidence couldn't be determined

This value is important because:

• It helps you prioritize which issues to address first
• It indicates when a fit assessment might be borderline
• It can guide how aggressively you should apply the recommendations

For example, a high confidence level for an overfitting detection suggests you should definitely take steps to reduce model complexity, while a low confidence might suggest more cautious adjustments.

FitType

Gets or sets the type of fit detected for the model.

public FitType FitType { get; set; }

Property Value

FitType

A value from the FitType enumeration indicating the type of fit.

Remarks

This property specifies the type of fit detected for the model, such as good fit, underfitting, or overfitting. A good fit indicates that the model appropriately captures the patterns in the data without being too simple or too complex. Underfitting occurs when the model is too simple to capture the underlying patterns in the data, resulting in poor performance on both training and test data. Overfitting occurs when the model is too complex and captures noise in the training data, resulting in good performance on training data but poor performance on test data. Other fit types might include multicollinearity (when predictor variables are highly correlated) or heteroscedasticity (when the variance of errors varies across the range of a predictor).

For Beginners: This property tells you what kind of fit was detected for your model.

The fit type:

• Identifies the relationship between your model and your data
• Helps you understand if your model needs to be adjusted
• Guides your next steps in the modeling process

Common fit types include:

• Good Fit: Your model appropriately captures the patterns in the data
• Underfitting: Your model is too simple and misses important patterns
• Overfitting: Your model is too complex and captures noise along with patterns
• Multicollinearity: Your predictor variables are too highly correlated
• Heteroscedasticity: The variance of errors changes across your data

For example, if this property is set to FitType.Overfitting, it means your model is likely too complex and needs to be simplified to generalize better.

Recommendations

Gets or sets a list of recommendations for improving the model fit.

public List<string> Recommendations { get; set; }

Property Value

List<string>

A list of string recommendations, initialized as an empty list.

Remarks

This property provides specific recommendations for improving the model fit based on the detected fit type. These recommendations are actionable suggestions that can help address issues with the model. For example, if overfitting is detected, recommendations might include adding regularization, reducing model complexity, or collecting more training data. If underfitting is detected, recommendations might include increasing model complexity, adding more features, or using a different type of model. The recommendations are provided as human-readable strings that can be directly presented to the user or logged for later review.

For Beginners: This list contains specific suggestions for improving your model.

The recommendations:

• Provide actionable steps to address the detected fit issues
• Are tailored to the specific fit type that was detected
• Can be directly presented to users or logged for later review

Common recommendations might include:

• For overfitting: "Add regularization to reduce model complexity"
• For underfitting: "Try a more complex model architecture"
• For multicollinearity: "Remove highly correlated features"

These recommendations are valuable because:

• They translate the technical diagnosis into practical actions
• They guide you toward better model performance
• They help you learn about best practices in modeling

For example, if your model is overfitting, the recommendations might include "Add L2 regularization" and "Increase dropout rate to 0.5".