Function Finder & Quick Reference

Find the right function for your task. All 60+ AnoFox Forecast functions organized by category.

Quick Function Finder

What I Need	Function	Category
Generate a forecast	`TS_FORECAST`	Forecasting
Forecast multiple series	`TS_FORECAST_BY`	Forecasting
Detect seasonality	`TS_DETECT_SEASONALITY`	Diagnostics
Find seasonal periods	`TS_DETECT_SEASONALITY_ALL`	Diagnostics
Detect structural breaks	`TS_DETECT_CHANGEPOINTS`	Diagnostics
Decompose series	`TS_STL_DECOMPOSITION`	Diagnostics
Multiple seasonality	`TS_MSTL_DECOMPOSITION`	Diagnostics
Fill time gaps	`TS_FILL_GAPS`	Data Preparation
Remove outliers	`TS_REMOVE_OUTLIERS`	Data Preparation
Extract features	`TS_TSFRESH_FEATURES`	Feature Engineering
Create lag features	`TS_LAG`	Feature Engineering
Rolling statistics	`TS_ROLLING_MEAN`	Feature Engineering
Fourier terms	`TS_FOURIER`	Feature Engineering
Measure accuracy	`TS_MAE`, `TS_RMSE`, `TS_MAPE`	Metrics
Check coverage	`TS_COVERAGE`	Metrics
Data quality report	`TS_QUALITY_REPORT`	Utilities

All Functions by Category

Forecasting Models (31)

AutoML Models

TS_FORECAST(..., 'AutoETS', ...)
TS_FORECAST(..., 'AutoARIMA', ...)
TS_FORECAST(..., 'AutoMFLES', ...)
TS_FORECAST(..., 'AutoMSTL', ...)
TS_FORECAST(..., 'AutoTBATS', ...)
TS_FORECAST(..., 'AutoTheta', ...)

Statistical Models

TS_FORECAST(..., 'ETS', ...) - Exponential Smoothing
TS_FORECAST(..., 'ARIMA', ...) - ARIMA
TS_FORECAST(..., 'Theta', ...) - Theta model
TS_FORECAST(..., 'Holt', ...) - Holt linear trend
TS_FORECAST(..., 'HoltWinters', ...) - Holt-Winters
TS_FORECAST(..., 'SES', ...) - Simple Exponential Smoothing

Advanced Models

TS_FORECAST(..., 'TBATS', ...) - Multiple seasonality
TS_FORECAST(..., 'MSTL', ...) - Multiple Seasonal & Trend
TS_FORECAST(..., 'MFLES', ...) - Robust to outliers
TS_FORECAST(..., 'GARCH', ...) - Volatility modeling

Baseline Models

TS_FORECAST(..., 'Naive', ...) - Last value
TS_FORECAST(..., 'SeasonalNaive', ...) - Last year
TS_FORECAST(..., 'RandomWalkDrift', ...) - Trend
TS_FORECAST(..., 'SeasonalWindowAverage', ...) - Average seasonal
TS_FORECAST(..., 'SMA', ...) - Simple Moving Average

Intermittent Demand

TS_FORECAST(..., 'Croston', ...) - Sparse demand
TS_FORECAST(..., 'CrostonOptimized', ...) - Optimized Croston
TS_FORECAST(..., 'CrostonSBA', ...) - Syntetos-Boylan
TS_FORECAST(..., 'ADIDA', ...) - Complex intermittency
TS_FORECAST(..., 'IMAPA', ...) - Multiple aggregation
TS_FORECAST(..., 'TSB', ...) - Trend + intermittent

Specialized

TS_FORECAST(..., 'DTW', ...) - Dynamic Time Warping
TS_FORECAST(..., 'AdIDA', ...) - Aggregate-Disaggregate
TS_FORECAST(..., 'Ensemble', ...) - Combine models

Diagnostics (8)

-- Seasonality Detection
TS_DETECT_SEASONALITY(table, date_col, value_col, params)
TS_DETECT_SEASONALITY_ALL(table, date_col, value_col, params)

-- Pattern Detection
TS_DETECT_CHANGEPOINTS(table, date_col, value_col, params)

-- Decomposition
TS_STL_DECOMPOSITION(table, date_col, value_col, params)
TS_MSTL_DECOMPOSITION(table, date_col, value_col, params)

-- Outlier Detection
TS_DETECT_OUTLIERS(table, value_col, params)
TS_DETECT_ANOMALIES(table, date_col, value_col, params)

Feature Engineering (15+)

-- Lag Features
TS_LAG(column, n_lags)

-- Rolling Statistics
TS_ROLLING_MEAN(column, window)
TS_ROLLING_STD(column, window)
TS_ROLLING_MIN(column, window)
TS_ROLLING_MAX(column, window)

-- Seasonal Features
TS_FOURIER(date, period, K)
TS_SEASONAL_DUMMY(date, period)

-- Trend Features
TS_TREND(column)
TS_DETREND(column)

-- Statistical Features
TS_TSFRESH_FEATURES(table, date_col, value_col, params)
TS_AUTOCORR(column, lag)

-- Advanced
TS_DIFFERENCE(column)
TS_LOG_TRANSFORM(column)

Evaluation Metrics (12)

-- Error Metrics
TS_MAE(actual_list, predicted_list)
TS_RMSE(actual_list, predicted_list)
TS_MAPE(actual_list, predicted_list)
TS_SMAPE(actual_list, predicted_list)
TS_MASE(actual_list, predicted_list, seasonal_period)

-- Bias Metrics
TS_ME(actual_list, predicted_list)
TS_BIAS(actual_list, predicted_list)

-- Goodness of Fit
TS_R_SQUARED(actual_list, predicted_list)

-- Interval Validation
TS_COVERAGE(actual_list, lower_list, upper_list)

-- Directional Accuracy
TS_DIRECTIONAL_ACCURACY(actual_list, predicted_list)

Data Preparation (12)

-- Cleaning
TS_FILL_GAPS(table, date_col, value_col, params)
TS_REMOVE_OUTLIERS(table, value_col, params)
TS_HANDLE_MISSING(table, date_col, value_col, params)

-- Quality Checks
TS_QUALITY_REPORT(table, date_col, value_col)
TS_VALIDATE_DATA(table, date_col, value_col, params)
TS_DETECT_DUPLICATES(table, date_col)

-- Transformation
TS_DESEASONALIZE(table, date_col, value_col, params)
TS_NORMALIZE(table, value_col, params)
TS_SCALE(table, value_col, params)

-- Aggregation
TS_AGGREGATE(table, date_col, value_col, frequency)
TS_RESAMPLE(table, date_col, value_col, new_frequency)

Multi-Series (GROUP BY)

All functions support GROUP BY for parallel processing:

-- Single series
SELECT * FROM TS_FORECAST('sales', 'date', 'amount', 'AutoETS', 30, {});

-- Multiple series (parallel)
SELECT * FROM TS_FORECAST_BY(
    'sales',
    'product_id',    -- GROUP BY column
    'date',
    'amount',
    'AutoETS',
    30,
    {}
);

Function Signatures

Core Forecasting

-- Single series forecast
TS_FORECAST(
    table_name: VARCHAR,
    date_col: VARCHAR,
    value_col: VARCHAR,
    model: VARCHAR,
    horizon: INT,
    params: MAP
) → TABLE(
    forecast_step INT,
    date_col DATE/TIMESTAMP,
    point_forecast DOUBLE,
    lower_X DOUBLE,
    upper_X DOUBLE,
    model_name VARCHAR
)

-- Multiple series (with GROUP BY)
TS_FORECAST_BY(
    table_name: VARCHAR,
    group_col: VARCHAR,
    date_col: VARCHAR,
    value_col: VARCHAR,
    model: VARCHAR,
    horizon: INT,
    params: MAP
) → TABLE(
    group_col_name VARCHAR,
    forecast_step INT,
    date_col DATE/TIMESTAMP,
    point_forecast DOUBLE,
    model_name VARCHAR
)

Common Parameters

Most functions accept a params MAP with optional settings:

-- Forecasting parameters
{
    'seasonal_period': 7,           -- For seasonal models
    'confidence_level': 0.95,       -- Prediction interval width
    'return_insample': false,       -- Include fitted values
    'trend': 'linear',              -- Trend type
    'method': 'auto'                -- Auto or manual selection
}

-- Detection parameters
{
    'method': 'auto',               -- Detection method
    'max_periods': 5,               -- Maximum seasonal periods to detect
    'hazard_lambda': 250            -- Changepoint sensitivity
}

-- Data parameters
{
    'max_gap': 7,                   -- Maximum gap to fill (days)
    'multiplier': 3.0,              -- Outlier threshold (std devs)
    'method': 'linear'              -- Imputation: linear, forward_fill, mean
}

Model Selection Guide

By Data Characteristics

Clean, trending data

✅ AutoETS
✅ Theta

Seasonal data (single period)

✅ AutoETS
✅ Holt-Winters

Multiple seasonality (hourly + daily + yearly)

✅ TBATS
✅ MSTL

Noisy, real-world data

✅ MFLES

Sparse, intermittent demand

✅ Croston
✅ ADIDA

Short-term forecasting

✅ Theta
✅ SeasonalNaive

Long-term forecasting

✅ AutoARIMA
✅ TBATS

Limited data (< 60 observations)

✅ SeasonalNaive
✅ Theta

Typical Workflow

1. Data Exploration
   ↓
   TS_DETECT_SEASONALITY() - Find patterns
   TS_DETECT_CHANGEPOINTS() - Find breaks
   TS_STL_DECOMPOSITION() - Understand components

2. Data Preparation
   ↓
   TS_FILL_GAPS() - Handle missing dates
   TS_REMOVE_OUTLIERS() - Clean anomalies
   TS_QUALITY_REPORT() - Validate data

3. Feature Engineering
   ↓
   TS_LAG() - Create lag features
   TS_ROLLING_MEAN() - Rolling statistics
   TS_TSFRESH_FEATURES() - Auto feature extraction

4. Forecasting
   ↓
   TS_FORECAST() - Generate predictions
   TS_FORECAST_BY() - Multiple series in parallel

5. Evaluation
   ↓
   TS_MAE() - Measure accuracy
   TS_COVERAGE() - Check intervals
   TS_MAPE() - Percentage error

6. Monitoring
   ↓
   Weekly: TS_QUALITY_REPORT()
   Monthly: Retrain with TS_FORECAST()

Performance Characteristics

Model	Speed	Memory	Best For
Naive	⚡⚡⚡	Minimal	Baseline
SeasonalNaive	⚡⚡⚡	Minimal	Simple seasonal
ETS	⚡⚡	Low	Trending + seasonal
Theta	⚡⚡	Low	Short-term
Holt-Winters	⚡⚡	Low	Clear patterns
AutoETS	⚡⚡	Low	Automatic tuning
ARIMA	🟢	Medium	Complex patterns
AutoARIMA	🟡	Medium	Auto parameter selection
TBATS	🟡	Medium	Multiple seasonality
MSTL	🟡	Medium	Very complex seasonality
MFLES	🟢	Medium	Noisy data, outliers
Croston	⚡⚡	Low	Sparse demand
DTW	🟡	High	Similarity-based

Legend: ⚡⚡⚡ Instant | ⚡⚡ Fast | 🟢 Medium | 🟡 Slow

Example: Complete Query

-- Comprehensive forecast with all features
WITH prepared AS (
    SELECT * FROM TS_FILL_GAPS('raw_sales', 'date', 'sales', {'method': 'linear'})
),
patterns AS (
    SELECT 'Seasonality' as metric, is_seasonal::VARCHAR as value FROM TS_DETECT_SEASONALITY('prepared', 'date', 'sales', {})
),
forecast AS (
    SELECT * FROM TS_FORECAST(
        'prepared',
        'date',
        'sales',
        'AutoETS',
        30,
        {'seasonal_period': 7, 'confidence_level': 0.95}
    )
)
SELECT
    forecast_step,
    date_col as forecast_date,
    ROUND(point_forecast, 2) as forecast,
    ROUND(lower_95, 2) as lower,
    ROUND(upper_95, 2) as upper
FROM forecast
WHERE forecast_step <= 7
ORDER BY forecast_step;

Next Steps

Choose a category to dive deeper:

Forecasting Models — All 31 models detailed
Feature Engineering — Lag, rolling, Fourier features
Diagnostics — Seasonality, changepoints, outliers
Metrics — 12 evaluation metrics explained
Utilities — Data prep and helper functions

API Cheat Sheet

-- Quick copy-paste templates

-- 1. Simple forecast
SELECT * FROM TS_FORECAST('table', 'date', 'value', 'AutoETS', 30, {});

-- 2. With seasonality
SELECT * FROM TS_FORECAST('table', 'date', 'value', 'AutoETS', 30, {'seasonal_period': 7});

-- 3. Multiple series
SELECT * FROM TS_FORECAST_BY('table', 'group_id', 'date', 'value', 'AutoETS', 30, {});

-- 4. With metrics
SELECT
    TS_MAE(LIST(actual), LIST(forecast)) as MAE,
    TS_RMSE(LIST(actual), LIST(forecast)) as RMSE,
    TS_MAPE(LIST(actual), LIST(forecast)) as MAPE
FROM comparison;

-- 5. Pattern detection
SELECT * FROM TS_DETECT_SEASONALITY('table', 'date', 'value', {});

-- 6. Data quality
SELECT * FROM TS_QUALITY_REPORT('table', 'date', 'value');

Troubleshooting Quick Ref

Problem	Solution
"Not enough data"	Need ≥60 observations, prefer ≥2 seasonal cycles
"Unknown model"	Check spelling: 'AutoETS' not 'autoets'
"Null parameter"	All params required: provide empty `{}` if unused
"Wrong output columns"	Different models return different columns
"Slow forecasting"	Use faster model (Naive, Theta) or TS_FORECAST_BY for parallelization
"Forecast all zeros"	Data too sparse → use Croston or ADIDA
"No seasonality detected"	May not exist, or need more data

Key Takeaways

✅ 31 forecasting models for different situations
✅ 8+ diagnostic functions for pattern detection
✅ 12+ feature engineering functions
✅ 12 evaluation metrics
✅ Support for millions of series with GROUP BY
✅ All functions documented with examples
✅ Use Function Finder table above to start

Quick Function Finder​

All Functions by Category​

Forecasting Models (31)​

AutoML Models​

Statistical Models​

Advanced Models​

Baseline Models​

Intermittent Demand​

Specialized​

Diagnostics (8)​

Feature Engineering (15+)​

Evaluation Metrics (12)​

Data Preparation (12)​

Multi-Series (GROUP BY)​

Function Signatures​

Core Forecasting​

Common Parameters​

Model Selection Guide​

By Data Characteristics​

Typical Workflow​

Performance Characteristics​

Example: Complete Query​

Next Steps​

API Cheat Sheet​

Troubleshooting Quick Ref​

Key Takeaways​