Forecasting (1): What is forecasting?

Summary

In this video series, Munem introduces the concept of forecasting, starting with basic models like moving averages, exponential smoothing, and regression using Excel. He explains that forecasts, while often inaccurate, provide valuable insights when done correctly. The video emphasizes the importance of forecasting in business, showing how those who forecast better perform better. Munem also illustrates what makes a good forecast, highlighting the significance of both training and test samples in determining forecast accuracy. The series will later explore more complex models like ARIMA, VAR, and machine learning techniques, using R software.

Takeaways

• 😀 Forecasting aims to predict future events, though forecasts are always subject to error.
• 😀 The first part of the video series covers basic forecasting models like moving averages, exponential smoothing, and Holt-Winters models using Excel.
• 😀 Advanced forecasting techniques, including ARIMA, VAR, machine learning, and neural networks, will be covered in later videos using R software.
• 😀 According to the Oxford English Dictionary, forecasting involves planning or predicting future events, but it is never fully accurate.
• 😀 Historical predictions, such as IBM’s 1943 statement about the limited market for computers, demonstrate the fallibility of forecasts.
• 😀 Business success often depends on the ability to make accurate forecasts and analyze data better than competitors.
• 😀 Those who can accurately forecast trends tend to perform better in business, as the ability to anticipate the future provides a competitive advantage.
• 😀 A good forecast model should fit both the training sample (historical data) and the test sample (future data).
• 😀 If a forecast works well on historical data but fails to predict future events, it is considered a poor forecast.
• 😀 Forecasting models can provide valuable insights even if they are not 100% accurate, helping businesses make more informed decisions.

Q & A

What is the primary focus of the video series that Munem is starting?

The video series will focus on forecasting, starting with simple forecasting models like Naïve forecast, Moving Average, Exponential Smoothing, and Holt-Winters Regression, followed by more complex models such as ARIMA, VAR, machine learning, and artificial neural networks.

Which software will be used for simple forecasting models in the first part of the series?

Excel will be used to demonstrate simple forecasting models in the first part of the video series.

What is the significance of forecasting according to the script?

Forecasting is important because it helps predict future events or behaviors, which can lead to better business decisions. Despite being imperfect, forecasts provide valuable insights that can lead to improved performance in business.

How does the Oxford English Dictionary define 'forecast'?

The Oxford English Dictionary defines 'forecast' as to arrange, plan, or predict something beforehand, essentially aiming to predict or determine the future.

What are some examples provided in the script where forecasts were wrong?

Two examples of inaccurate forecasts were provided: one from 1943, where the chairman of IBM predicted a market for only five computers, and another from the 1950s, where the US president claimed no one would want a computer in their home.

What is the key takeaway from the examples of wrong forecasts mentioned in the script?

The key takeaway is that forecasts are often inaccurate, but they can still provide valuable insights if done correctly, and are crucial for better decision-making in business.

Why is good forecasting important in business?

Good forecasting is important in business because companies that forecast well tend to perform better. Those who can analyze data and predict trends more accurately have a competitive advantage over those who cannot.

What is the difference between a training sample and a test sample in forecasting?

A training sample is used to build the forecasting model, while a test sample is used to evaluate how well the model performs on unseen data. A good forecast should perform well in both the training and test samples.

In the example given, why is the forecast considered poor after week 40?

The forecast is considered poor after week 40 because it works well in the training sample (before week 40) but fails to predict accurately for the test sample (after week 40), indicating that the model doesn't generalize well to new data.

What will be covered in the second part of the video series?

In the second part of the video series, more complex forecasting models such as ARIMA, VAR, machine learning, and artificial neural networks will be introduced, and R software will primarily be used for these models.