Group Overview

Collective performance across all members · forecasting_dashboard.csv

TOTAL GROUP BALANCE

—

latest closing period

TOTAL CONTRIBUTIONS

—

all periods combined

TOTAL WITHDRAWALS

—

rate: —

ANOMALY EVENTS

—

— clean periods

Group balance over time

Balance by member latest period

Monthly net flowcontrib − withdraw

Member categories

Anomalies per period

Member summary 24-month aggregate · forecasting_dashboard.csv

MEMBER	REGION	CATEGORY	CLOSING BALANCE	TOTAL CONTRIB	TOTAL WITHDRAW	ANOMALIES	BEST MODEL	AVG MAPE

Member Deep-dive

Balance trajectory · actual vs forecast · 2026 transactions · anomaly log

Balance trajectory

Actual Forecast Anomaly

Contributions vs Withdrawals

Anomaly log —

DATE	TYPE	BALANCE	WITHDRAWAL

2026 Transaction history stokvel_dataset.csv

DATE	FREQUENCY	CONTRIBUTION	WITHDRAWAL	BALANCE

Forecast Accuracy

Holt-Winters vs ARIMA · RMSE · MAE · MAPE per member

HW MODEL WINS

—

lower avg RMSE

BEST HW RMSE

—

across all periods

GROUP AVG MAPE

—

across all members

WORST ARIMA RMSE

—

across all periods

RMSE — Holt-Winters vs ARIMA

Holt-Winters ARIMA

Best model per member

MAPE over time — all members%

Average MAE per member

Full accuracy table

MEMBER	DATE	RMSE HW	RMSE ARIMA	MAE	MAPE	BEST MODEL	ACTUAL	FORECAST	ERROR

Anomaly Tracker

Rule-based detection · Sudden Drop · Spike · all flagged events

ℹ

No ML model on anomaly detection. All flags are deterministic threshold rules inside scripts/forecasting.py — Sudden Drop (withdrawal >30% balance drop vs previous period) and Spike (large single withdrawal). Consider adding Isolation Forest or rolling Z-score for statistical detection.

TOTAL EVENTS

—

anomaly rate: —

SUDDEN DROPS

—

withdrawal > 30% balance drop

SPIKES

—

large single withdrawal

CLEAN PERIODS

—

no anomaly flagged

Anomalies by period

Sudden Drop Spike

Count per member

Event log

MEMBER	DATE	TYPE	BALANCE	WITHDRAWAL	CONTRIBUTION	REGION

Risk summary

Regional View

Cape Town vs Durban · by category · contribution frequency breakdown

REGION 1

—

REGION 2

—

Balance by region over time

Contributions by region

Anomalies by region

Balance by category

Forecast horizon split

Contribution frequency (2026 data) stokvel_dataset.csv

Regional performance table

MEMBER	REGION	CATEGORY	HORIZON	FREQUENCY	CLOSING BAL	AVG CONTRIB	AVG WITHDRAW	ANOMALIES	AVG RMSE HW	AVG RMSE AR