Membandingkan Grey Forecasting dengan Single Eksponensial Smoothing (SES)

Summary

In this presentation, Dinar Olivia Rahmadita explores the comparative analysis between Gray Forecasting (GM 1.1) and Single Exponential Smoothing (SES) for predicting air humidity in Jakarta. Gray forecasting is suitable for small datasets and uncertainty, while SES is best for stable data without trends. The analysis reveals that SES outperforms Gray Forecasting in terms of accuracy with a lower MAPE of 2.44%, compared to 9.86% for GM 1.1. The video emphasizes the strengths and weaknesses of both methods and concludes that SES is more effective for short-term predictions, while Gray Forecasting is useful for limited data scenarios.

Takeaways

• 😀 Gray forecasting is used for systems with partial information, where neither the data is completely known (white) nor completely unknown (black).
• 😀 The GM 1.1 model is the most common in gray forecasting, focusing on one variable and involving steps like parameter estimation and prediction equations.
• 😀 Gray forecasting is suitable for small datasets (less than 20 points), particularly for forecasting demand, new products, price predictions, and population data.
• 😀 Gray forecasting advantages include no assumption of stable data distribution and ease of implementation, but it struggles with fluctuating data and long-term accuracy.
• 😀 Single exponential smoothing (SES) is a time series forecasting method for data without trends or seasonality, giving more weight to recent data.
• 😀 SES is suitable for stable data and is quick and easy to calculate, but it’s not effective for data with trends or seasonality and is highly dependent on alpha selection.
• 😀 The Jakarta air humidity data from 2016 to 2025 was analyzed using the ADF test, showing that the data became stationary after first differencing.
• 😀 For the gray forecasting (GM 1.1), the MAPE for one-period prediction was 9.86%, while for SES, it was significantly lower at 2.44%.
• 😀 The predicted values for one period ahead were similar for both methods: gray forecasting predicted 77.902, and SES predicted 77.54, indicating consistency.
• 😀 While gray forecasting is useful for short-term data with uncertainty and noise, SES is more accurate for stable data with minimal trend or seasonality.
• 😀 The conclusion recommends SES for short-term prediction accuracy in the Jakarta humidity data, as its lower MAPE (2.44%) outperforms gray forecasting for this particular dataset.

Q & A

What is Grey Forecasting, and why is it called 'grey'?

Grey Forecasting is used to analyze systems with partial or incomplete information, not completely known (white) or completely unknown (black). It is called 'grey' because it deals with uncertainties or incomplete data, filling the gap between the extremes of certainty and ignorance.

What is the most common model used in Grey Forecasting?

The most common model in Grey Forecasting is GM 1.1, which stands for 'Grey' forecasting, 'Model 1', and '1' indicating a first-order model with a single variable.

What are the key steps involved in Grey Forecasting?

The key steps in Grey Forecasting include accumulation generating operation, differential model GM 1.1 parameter estimation, and prediction equation for original value prediction.

What are the advantages of using Grey Forecasting?

Grey Forecasting is advantageous because it doesn't require assumptions about stable data distribution and is effective for small, incomplete datasets. It’s especially useful when data is insufficient or noisy.

What are the disadvantages of Grey Forecasting?

The disadvantages of Grey Forecasting include its ineffectiveness with highly fluctuating data, a decrease in accuracy over the long term, and lower performance compared to statistical methods for big data.

How does Single Exponential Smoothing (SES) work?

Single Exponential Smoothing (SES) is a time series forecasting method where more weight is given to the most recent data. The model uses an exponential weighting system to predict future values based on the smoothed past data.

What is the main limitation of Single Exponential Smoothing (SES)?

The main limitation of SES is that it is not effective for data that has trends or seasonality. It performs best with relatively stable data without significant changes over time.

What does the MAPE (Mean Absolute Percentage Error) tell us in forecasting?

MAPE measures the average error in percentage terms between the forecasted and actual values. A smaller MAPE indicates a more accurate forecasting model, with values less than 5% being considered very good.

Which forecasting method performed better based on MAPE in the analysis?

Based on MAPE, Single Exponential Smoothing (SES) performed better with a MAPE of 2.44%, compared to Grey Forecasting's 9.86%, indicating that SES was more accurate in predicting the data.

What were the forecasted values for the next period using both forecasting methods?

For the next period, the forecasted value using Grey Forecasting was 77.902, while Single Exponential Smoothing predicted a value of 77.54. Both values were close, indicating consistency in capturing data trends.

When should you prefer to use Grey Forecasting over Single Exponential Smoothing?

You should prefer Grey Forecasting when the dataset is small, incomplete, or when historical data is not fully available. It is particularly useful in scenarios with limited or uncertain data.

What is the final recommendation based on the analysis for short-term forecasting accuracy?

For short-term forecasting accuracy, Single Exponential Smoothing (SES) is recommended due to its lower MAPE of 2.44%, indicating its superior accuracy for this humidity data.