Class ConditionalInferenceTreeOptions
Configuration options for Conditional Inference Trees, a statistically-driven approach to decision tree learning.
public class ConditionalInferenceTreeOptions : DecisionTreeOptions- Inheritance
-
ConditionalInferenceTreeOptions
- Inherited Members
Remarks
Conditional Inference Trees use statistical tests to select variables and determine split points, which helps reduce selection bias toward variables with many possible split points. This approach often produces more reliable and statistically sound trees compared to traditional methods.
For Beginners: Conditional Inference Trees are a special type of decision tree that uses statistics to make better decisions about how to split data. Regular decision trees sometimes favor certain types of data unfairly (like preferring variables with more possible values). This approach is like having a referee that makes sure the tree-building process is fair and statistically sound. The result is often a more reliable model, especially for data where some variables have many possible values and others have few.
Properties
MaxDegreeOfParallelism
Gets or sets the maximum number of parallel operations when building the tree.
public int MaxDegreeOfParallelism { get; set; }Property Value
- int
The maximum degree of parallelism, defaulting to the number of processor cores available.
Remarks
This setting controls how many CPU cores the algorithm will use when building the tree. Higher values can speed up training on multi-core systems but may increase memory usage. The default uses all available processor cores for maximum performance.
For Beginners: This setting controls how many parts of the tree-building process can happen at the same time. By default, it uses all the processing power your computer has available (all CPU cores). Think of it like having multiple workers building different branches of the tree simultaneously. This makes the process faster, but sometimes uses more memory. If your computer is struggling with memory issues when building large trees, you might want to reduce this number (e.g., to half the number of cores). For most users, the default setting works well and provides the best performance.
SignificanceLevel
Gets or sets the statistical significance level used for hypothesis testing when selecting split variables.
public double SignificanceLevel { get; set; }Property Value
- double
The significance level as a decimal between 0 and 1, defaulting to 0.05 (5%).
Remarks
The significance level determines how strong the evidence must be before the algorithm will split on a variable. Lower values (e.g., 0.01) require stronger evidence and typically result in simpler trees. Higher values (e.g., 0.1) accept weaker evidence and may produce more complex trees.
For Beginners: This setting controls how "picky" the algorithm is when deciding whether to split the data. The default value (0.05 or 5%) is a standard threshold in statistics. Think of it like a judge in a courtroom - a strict judge (lower value like 0.01) needs very strong evidence before making a decision, resulting in fewer splits and a simpler tree. A more lenient judge (higher value like 0.1) will accept weaker evidence, potentially creating more splits and a more complex tree. If your tree seems too simple, you might increase this value slightly; if it's too complex, you might decrease it.
StatisticalTest
Gets or sets the type of statistical test used to evaluate potential splits.
public TestStatisticType StatisticalTest { get; set; }Property Value
- TestStatisticType
The statistical test type, defaulting to TTest.
Remarks
Different statistical tests are appropriate for different types of data: - T-tests are suitable for continuous target variables - Chi-square tests work well for categorical target variables - ANOVA is useful when comparing multiple groups The algorithm will automatically select an appropriate test in many cases, but this option allows manual override.
For Beginners: This setting determines what kind of statistical method the algorithm uses to decide if a split is meaningful. Think of it like choosing the right tool for a job - you wouldn't use a hammer to tighten a screw. The default (TTest) works well for numerical data (like predicting prices or temperatures). If you're working with categorical data (like predicting categories or yes/no outcomes), a different test like ChiSquare might be more appropriate. In many cases, the algorithm can select the right test automatically, so you only need to change this if you have specific knowledge about your data.