Prediktori

Prediktori is a term used in statistics and data analysis to denote variables that are used to predict a target outcome. In predictive modeling, prediktori serve as inputs or features that carry information about the quantity being estimated. The concept applies across diverse methods, including linear and logistic regression, decision trees, and machine learning algorithms.

Etymology and usage: The word prediktori derives from the same root as predictor, reflecting its role in foreseeing or estimating values. In many languages, prediktori is the standard label for input variables in a model, distinguishing them from the dependent or outcome variable.

Role in modeling: Prediktori are selected to explain or forecast the target variable. They can be quantitative (such as age or income), categorical (such as region or category), or derived through feature engineering (such as interaction terms or transformed scales). The choice and quality of prediktori affect model performance, interpretability, and generalization to new data.

Data considerations: Good prediktori should be relevant, measured accurately, and free from leakage of information about the target variable. Problems such as multicollinearity, missing values, and measurement error can reduce predictive accuracy and complicate interpretation.

Examples: In health research, prediktori for a cardiovascular risk score might include blood pressure, cholesterol level, and smoking status. In marketing, prediktori for customer churn could include usage frequency, tenure, and customer satisfaction scores.

Limitations: Prediktori are tools for prediction, not proof of causation. Correlation between a prediktori and the target does not imply a causal relationship, and overfitting can arise if too many or overly complex prediktori are used. See also predictor, feature, regression, and machine learning.