Forecasting

1. Polynomial Trends - A mathematical model where the trend is represented by a polynomial function of time, typically in the form yt=β0+β1t+β2t2+⋯+βptpy_t = \beta_0 + \beta_1 t + \beta_2 t^2 + \dots + \beta_p t^pyt​=β0​+β1​t+β2​t2+⋯+βp​tp.

2. Forecasting with Polynomial Trends - To forecast beyond the sample, you plug in the future time value T+jT+jT+j into the polynomial equation yT+j=β^0+β^1(T+j)+⋯+β^p(T+j)py_{T+j} = \hat{\beta}_0 + \hat{\beta}_1 (T+j) + \dots + \hat{\beta}_p (T+j)^pyT+j​=β^​0​+β^​1​(T+j)+⋯+β^​p​(T+j)p.

3. Power Series Expansion - A method in mathematics that expresses functions as an infinite sum of terms involving powers of the argument; a polynomial can be a truncated version of this series.

4. Exponential Growth - A process where the growth rate is proportional to the current value, modeled as yt=β0eβ1ty_t = \beta_0 e^{\beta_1 t}yt​=β0​eβ1​t, often used in financial contexts.

5. Compound Interest - Exponential growth in finance, where the value of an investment grows according to the formula FV=PV(1+r)tFV = PV (1 + r)^tFV=PV(1+r)t, with rrr being the annual interest rate.

6. Logistic Growth - A model that incorporates an upper limit to growth, represented by yt=γ1+eβ0+β1ty_t = \frac{\gamma}{1 + e^{\beta_0 + \beta_1 t}}yt​=1+eβ0​+β1​tγ​, commonly used when growth is limited by factors like resources.

7. Saturation Point - In logistic growth, the saturation point is where the growth rate decreases and the system approaches its upper bound or maximum value.

8. Exponential vs. Logistic Growth - Exponential growth assumes no limits, while logistic growth accounts for constraints and slows as it nears an upper limit.

9. Holt’s Two-Parameter Model - A forecasting method that accounts for both level and trend by using two smoothing parameters, α\alphaα (level) and β\betaβ (trend).

10. Holt’s Model Forecast Equation - The forecast for time t+pt+pt+p is given by yt+p=Ft+pTty_{t+p} = F_t + pT_tyt+p​=Ft​+pTt​, where FtF_tFt​ is the level at time ttt and TtT_tTt​ is the trend at time ttt.

11. Exponential Smoothing - A forecasting technique where weights decrease exponentially for past observations, with more recent data given higher weights.

12. Level Smoothing Equation in Holt’s Model - Ft=αyt+(1−α)(Ft−1+Tt−1)F_t = \alpha y_t + (1 - \alpha)(F_{t-1} + T_{t-1})Ft​=αyt​+(1−α)(Ft−1​+Tt−1​), where FtF_tFt​ is the smoothed value, and Tt−1T_{t-1}Tt−1​ is the previous trend.

13. Trend Smoothing Equation in Holt’s Model - Tt=β(Ft−Ft−1)+(1−β)Tt−1T_t = \beta(F_t - F_{t-1}) + (1 - \beta)T_{t-1}Tt​=β(Ft​−Ft−1​)+(1−β)Tt−1​, where TtT_tTt​ is the smoothed trend value at time ttt.

14. Holt-Winters’ Seasonal Model - An extension of Holt’s method that includes a seasonal component to account for seasonality in time series data.

15. Additive Seasonal Model - A version of the Holt-Winters’ model where the seasonal component is added to the level and trend, suitable for constant seasonal variations.

16. Multiplicative Seasonal Model - A version of the Holt-Winters’ model where the seasonal component is multiplied by the level and trend, suitable for changing seasonal variations.

17. Seasonal Component in Holt-Winters - Captures the repeating patterns or fluctuations within the data that recur at regular intervals, such as yearly or monthly cycles.

18. Holt-Winters Forecast Equation (Additive) - yt+p=Ft+pTt+Ct+p−ry_{t+p} = F_t + pT_t + C_{t+p-r}yt+p​=Ft​+pTt​+Ct+p−r​, where FtF_tFt​ is the level, TtT_tTt​ is the trend, and Ct+p−rC_{t+p-r}Ct+p−r​ is the seasonal component.

19. Initial Values in Holt-Winters - For initialization, the first level is set as F1=y1F_1 = y_1F1​=y1​, the first trend is T1=y2−y1T_1 = y_2 - y_1T1​=y2​−y1​, and the first seasonal component is C1=0C_1 = 0C1​=0.

20. Forecast Accuracy - Measures how well a forecasted value matches the actual observed value, with metrics like Mean Error (ME), Mean Absolute Deviation (MAD), and Root Mean Square Error (RMSE).

21. Mean Error (ME) - A measure of forecast bias, calculated as the average of the differences between the observed values and the forecasted values. It indicates whether the model consistently over- or under-forecast.

22. Mean Absolute Deviation (MAD) - A measure of forecast accuracy that averages the absolute differences between observed and forecasted values, providing an indication of the typical size of forecast errors.

23. Root Mean Square Error (RMSE) - A measure of forecast accuracy that calculates the square root of the average of squared forecast errors, emphasizing larger errors more than MAD.

24. Mean Absolute Percent Error (MAPE) - A relative measure of forecast accuracy that expresses errors as a percentage of the actual values, making it easier to compare accuracy across different datasets.

25. Bias in Forecasting - Occurs when the mean error (ME) is consistently different from zero, indicating that the model tends to over- or under-predict.

26. Ex-post Forecast Errors - Errors that are observed after the forecast is made, which are used to evaluate and improve the forecasting model’s performance.

27. Ex-ante Forecast Errors - Forecast errors made when predicting future observations based on past data, which are crucial for assessing how well a model can predict unknown data.

31. Pegel’s Classification - A framework for classifying exponential smoothing methods based on trend and seasonal components, including additive and multiplicative models.

32. Additive Model - A model where the seasonal variations are constant over time and the seasonal component is added to the level and trend.

33. Multiplicative Model - A model where the seasonal variations change proportionally with the level of the time series and are multiplied with the level and trend.

34. Trend Component - The underlying direction in the data over time, whether it is increasing, decreasing, or flat, often modeled in forecasting methods like Holt’s.

35. Smoothing Parameters α\alphaα and β\betaβ - Parameters used in exponential smoothing models; α\alphaα controls the smoothing of the level, and β\betaβ controls the smoothing of the trend.

36. Polynomial Trend - A mathematical model used to fit a curve to the time series data, represented by a polynomial equation yt=β0+β1t+β2t2+⋯+βptpy_t = \beta_0 + \beta_1 t + \beta_2 t^2 + \dots + \beta_p t^pyt​=β0​+β1​t+β2​t2+⋯+βp​tp, where ppp is the degree of the polynomial.

37. Degree of Polynomial ppp - The highest power of ttt in the polynomial trend equation, determining how complex the trend line is (higher degrees capture more complex patterns).

38. Forecast with Polynomial Trend - A forecast for future time periods using a polynomial equation, such as yT+j=β^0+β^1(T+j)+β^2(T+j)2+…y_{T+j} = \hat{\beta}_0 + \hat{\beta}_1 (T+j) + \hat{\beta}_2 (T+j)^2 + \dotsyT+j​=β^​0​+β^​1​(T+j)+β^​2​(T+j)2+…, where β^i\hat{\beta}_iβ^​i​ are estimated coefficients.

39. Exponential Growth - A model of growth where the value increases at a constant percentage rate over time, often used in financial and economic forecasting.

40. Exponential Growth Equation - yt=β0eβ1ty_t = \beta_0 e^{\beta_1 t}yt​=β0​eβ1​t, representing growth where yty_tyt​ increases exponentially with time ttt, with β0\beta_0β0​ as the initial value and β1\beta_1β1​ as the growth rate.

41. Logistic Growth - A type of growth where the growth rate slows down as it approaches a maximum value or carrying capacity, often used to model resource-limited growth.

42. Logistic Growth Equation - yt=γ1+eβ0+β1ty_t = \frac{\gamma}{1 + e^{\beta_0 + \beta_1 t}}yt​=1+eβ0​+β1​tγ​, where γ\gammaγ is the maximum value, and β0\beta_0β0​ and β1\beta_1β1​ are parameters controlling the growth rate.

43. S-Curve - A graph that depicts logistic growth, where the growth rate starts slowly, accelerates, and then decelerates as the value approaches a maximum capacity or limit.

44. Saturation Point - The point in logistic growth where the growth rate slows significantly as the system reaches its carrying capacity, commonly seen in markets or populations.

45. Upper Bound in Logistic Growth - The maximum level that can be reached in a logistic growth model, beyond which growth cannot continue due to resource constraints.

46. Forecasting Accuracy Measures - Metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) are used to evaluate how well a forecasting model performs.

47. Overfitting - A situation where a model fits the training data too well, capturing noise or random fluctuations, which leads to poor performance on new data.

48. Underfitting - A situation where a model is too simple to capture the underlying trends or patterns in the data, resulting in inaccurate forecasts.

49. Forecasting Horizon - The time period over which forecasts are made, typically broken into short-term, medium-term, and long-term horizons, each with different forecasting methods.

50. Model Validation - The process of assessing a forecasting model’s accuracy and reliability using historical data that was not part of the training set.

51. Holt’s Linear Exponential Smoothing - A forecasting method that accounts for both trend and level in a time series, using two parameters: α\alphaα for level smoothing and β\betaβ for trend smoothing.

52. Holt-Winters Method - An extension of Holt’s method that includes seasonality, with both additive and multiplicative seasonal components to capture trends, levels, and seasonal variations.

53. Additive Seasonal Model - A seasonal forecasting method where the seasonal variations are constant and are added to the forecast after adjusting for trend and level.

54. Multiplicative Seasonal Model - A seasonal forecasting method where seasonal variations are proportional to the level of the time series, adjusting by multiplying the seasonal component.

55. Seasonal Component StS_tSt​ - The repeating pattern in a time series, typically due to factors such as seasonality or cyclic changes, which can be modeled separately in forecasting.

56. Smoothing Equations in Holt-Winters - Equations that update the level, trend, and seasonal components over time in Holt-Winters' method, adjusting forecasts accordingly.

57. Alpha (α\alphaα) in Exponential Smoothing - The smoothing parameter that determines how much weight is given to the most recent observation in the level equation.

58. Beta (β\betaβ) in Exponential Smoothing - The smoothing parameter that adjusts the trend component of the forecast, determining how much influence the previous trend has on the forecast.

59. Gamma (γ\gammaγ) in Holt-Winters - The smoothing parameter that determines how much weight is given to the seasonal component when updating seasonal adjustments.

60. Stationary Series - A time series whose statistical properties such as mean and variance do not change over time, making it simpler to model and forecast.

61. Non-Stationary Series - A time series with changing statistical properties over time, often requiring transformations like differencing or detrending to make it stationary.

62. Time Series Decomposition - The process of breaking down a time series into components: trend, seasonal, and residual (or error) to better understand the underlying structure.

63. Moving Average (MA) - A forecasting technique where the forecast is the average of a specified number of past data points, useful for smoothing out short-term fluctuations.

64. Autoregressive Integrated Moving Average (ARIMA) - A class of models that combine autoregressive (AR), differencing (I), and moving average (MA) components to model time series data.

65. Forecast Error - The difference between the actual and forecasted values, used to evaluate the accuracy of a forecasting model.

66. Bias in Forecasting - Systematic error where forecasts consistently overestimate or underestimate the actual values, often caused by model assumptions or parameters.

67. Confidence Interval in Forecasting - A range of values within which the true value of the forecast is likely to fall, used to express uncertainty around a point forecast.

68. Model Overfitting - When a model captures noise or random fluctuations in the data rather than the true underlying pattern, reducing its ability to generalize to new data.

69. Model Underfitting - When a model is too simple and does not capture the complexity of the data, leading to inaccurate forecasts.