features

Documentation

Overview

Package features provides utilities for building feature frames from raw metric data.

Index

• func FillMissingValues(frame models.FeatureFrame) models.FeatureFrame
• type Builder
  • func NewBuilder() *Builder
  • func (b *Builder) BuildFeatures(df adapters.DataFrame) (models.FeatureFrame, error)

Functions

func FillMissingValues

func FillMissingValues(frame models.FeatureFrame) models.FeatureFrame

FillMissingValues fills missing values in a FeatureFrame using forward fill strategy. For each feature column, missing values (represented as NaN or not present) are replaced with the last valid value seen.

This is a simple implementation for v0.1. More sophisticated imputation strategies (mean, interpolation) can be added later if needed.

Types

type Builder

type Builder struct{}

Builder constructs feature frames from DataFrames, extracting time-based features and transforming raw metric data into a format suitable for forecasting models.

func NewBuilder

func NewBuilder() *Builder

NewBuilder creates a new feature builder.

func (*Builder) BuildFeatures

func (b *Builder) BuildFeatures(df adapters.DataFrame) (models.FeatureFrame, error)

BuildFeatures converts a DataFrame from an adapter into a FeatureFrame for a model. It extracts the following features from each row:

• value: the metric value (required)
• timestamp: Unix timestamp in seconds (extracted from "ts" field if present)
• hour: hour of day (0-23) extracted from timestamp
• day: day of week (0-6, Sunday=0) extracted from timestamp

Rows without a "value" field are skipped. If "ts" field is missing, features derived from timestamps are not included.