Forecasting & Replenishment. Forecasting Forecasting & Replenishment Historical Usage Forecasting.
Forecasting in HCSC
Transcript of Forecasting in HCSC
![Page 1: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/1.jpg)
Forecasting in HCSC
![Page 2: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/2.jpg)
Type of preditions
Predictions
Nu
mb
er o
f pa
tients
…relationship respect to other variables
…the historical data of the variable
![Page 3: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/3.jpg)
Formulating relationship
between decision variables
![Page 4: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/4.jpg)
• Pearson r Coefficient: measure the strength of the linear
association between continuous variables.
• Spearman rho Coefficient: is used to estimate the degree of
association between two ordinal variables measuring the
concordance or discordance.
• Kendall Coefficient: measures the strength of dependence
between two nominal variables with same scale.
Bivariate Analysis
Correlation
Coefficient (r)
Correlation coefficient varies from -1 (Negative Correlation) to +1 (Positive Correlation)
If Correlation coefficient is 0, there is NO Correlation or association
![Page 5: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/5.jpg)
• Pearson r Coefficient: measure the strength of the linear
association between continuous variables.
Bivariate Analysis
Correlation
Coefficient (r)
Correlation coefficient varies from -1 (Negative Correlation) to +1 (Positive Correlation)
If Correlation coefficient is 0, there is NO Correlation or association
𝑟𝑥𝑦 =σ𝑖=1𝑛 (𝑥𝑖 − ҧ𝑥) ∙ (𝑦𝑖 − ത𝑦)
σ𝑖=1𝑛 (𝑥𝑖 − ҧ𝑥) ∙ σ𝑖=1
𝑛 (𝑦𝑖 − ത𝑦)
Example in Excel…(Pills
consumption_dataset)
![Page 6: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/6.jpg)
YX
People will take more flu pills when the temperature growths?
Response variable or dependent variable
Independent variable
![Page 7: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/7.jpg)
Strong a positive relationship
Strong a positive relationship
REAL STATISTICS Add ins…
![Page 8: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/8.jpg)
Y X
… …
Is there any clear relationship…?
![Page 9: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/9.jpg)
Regression analysis…
![Page 10: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/10.jpg)
• Linear Regression: establishes a relationship between
dependent variable and one or more independent variables
using a best fit straight line (continuous).
• Logistics Regression: explain the relationship between
dependent binary variable and one or more (continuous)
independent variables.
• Nonlinear Regression: the relationship between the
dependent and independent parameters are not linear.
• Ordinal Regression: explains the relationship between a
dependent variable and independent variables (ordinal).
Multivariate Analysis
Regression
Analysis
Y= dependent variable
ß0 = Y intercept
ß1 = slope coefficient
X= independent variable
![Page 11: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/11.jpg)
𝑌 = 𝛽0 + 𝛽1𝑋
Regression equation
(useful for predictions)
Least-squares method
![Page 12: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/12.jpg)
YX
People will take more flu pills when the temperature growths?
Let's find the regression
equation…predictions…
Response variable or dependent variable
Independent variable
![Page 13: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/13.jpg)
Strong and positive relationship…
![Page 14: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/14.jpg)
H0: 𝑌 = 𝛽0 + 𝛽1𝑋 …𝛽1 ≈ 0H1: 𝑌 = 𝛽0 + 𝛽1𝑋 …𝛽1 ≠ 0
Result: Reject H0…regression is SIG
Rule:
If p-value <= Alpha (α) ThenReject H0
ElseAccept H0
![Page 15: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/15.jpg)
Equation: 𝑌 = −36.3612 + 3.7379𝑋Both coefficients contribute significantly
to explain the Y´s variability
Pill_consumption = -36.3612 + 3.7379 x MAX-Temperature
![Page 16: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/16.jpg)
Residual graphs
![Page 17: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/17.jpg)
Assumptions (simple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No-autocorrelation (residuals)
![Page 18: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/18.jpg)
Assumptions (simple-reg.):
❑ Linear relationship (OK)
❑ Y ~ Normal Distribution
❑ No-autocorrelation (residuals)
Clear linear relationship…
![Page 19: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/19.jpg)
Assumptions (simple-reg.):
❑ Linear relationship (OK)
❑ Y ~ Normal Distribution (OK)
❑ No-autocorrelation (residuals)
Anderson-Darling Test:
H0: the observed variable fits to a Normal DistributionH1: the observed variable does not fit to a Normal Distribution
P-value > 0.05 (user-defined parameter) …ThenAccept H0 (the dependent variable follows a Normal Distribution)
![Page 20: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/20.jpg)
Assumptions (simple-reg.):
❑ Linear relationship (OK)
❑ Y ~ Normal Distribution (OK)
❑ No-autocorrelation (residuals) (OK)
1.5 < D-stat < 2.5
The assumption is met
𝑑 =σ𝑖=2𝑛 (𝑒𝑖 − 𝑒𝑖−1)
2
σ𝑖=1𝑛 𝑒𝑖
2
No auto-correlation (residuals)
![Page 21: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/21.jpg)
Multiple regression…
![Page 22: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/22.jpg)
Multiple regression:
𝑌 = 𝛽𝑋 + 𝜀Model:
𝑌 =
𝑌1𝑌2...𝑌𝑛
𝑋 =
1 𝑋11 𝑋12 … 𝑋1𝑏1 𝑋21 𝑋22 … 𝑋2𝑏..
.
1
.
.
.
𝑋𝑛1
.
.
.
𝑋𝑛2
.
.
.
…
.
.
.
𝑋𝑛𝑝
𝛽 =
𝛽0𝛽1...𝛽𝑏
𝜀 =
𝜀1𝜀2...𝜀𝑛
?
http://www.stat.ucla.edu/~hqxu/stat105/pdf/ch12.pdf
![Page 23: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/23.jpg)
Multiple regression:
𝑌 = 𝛽𝑋 + 𝜀Model:
𝑌 =
𝑌1𝑌2...𝑌𝑛
𝑋 =
1 𝑋11 𝑋12 … 𝑋1𝑏1 𝑋21 𝑋22 … 𝑋2𝑏..
.
1
.
.
.
𝑋𝑛1
.
.
.
𝑋𝑛2
.
.
.
…
.
.
.
𝑋𝑛𝑝
𝛽 =
𝛽0𝛽1...𝛽𝑏
𝜀 =
𝜀1𝜀2...𝜀𝑛
መ𝛽 = (𝑋𝑇𝑋)−1𝑋𝑇𝑌Seeking the coefficients…
𝑌 = 𝑋 መ𝛽
![Page 24: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/24.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
![Page 25: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/25.jpg)
YX2X1Linear relationship…
![Page 26: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/26.jpg)
YX2X1Linear relationship…
![Page 27: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/27.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution (YES--- α = 0.01)
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
![Page 28: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/28.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)No significant correlation between
independent variables…OK
![Page 29: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/29.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
The variance are significantly different…
![Page 30: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/30.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
The variance are significantly different…
Z-score transformation = (x – μ) / σ
![Page 31: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/31.jpg)
Assumptions (multiple-reg.):
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
![Page 32: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/32.jpg)
Assumptions (multiple-reg.):
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals – normality as well)
Variance are now equal…OK
![Page 33: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/33.jpg)
Assumptions (multiple-reg.):
❑ Linear relationship
❑ Y ~ Normal Distribution
❑ No multicollinearity in the data (X-matrix)
❑ Homoscedasticity (X-matrix)
❑ No-autocorrelation (residuals)
You might assume that is OK…
![Page 34: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/34.jpg)
multiple-reg (REAL STATS-Add ins):
![Page 35: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/35.jpg)
multiple-reg:
Setting the Xs and Y
Setting the confidence level of the regression analysis
Run the whole Regression Analysis
Auto-correlation analysis…
![Page 36: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/36.jpg)
multiple-reg:
The two independent variables explain 76% of the variance in the
dependent variable
The regression equation presents a standard error of 16 patients…
The regression is significant…meani
ng: you can use the regression equation for predictions…
![Page 37: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/37.jpg)
multiple-reg:
Only the intercept and the relative humidity have a
significant influence on the
number of patients…
![Page 38: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/38.jpg)
multiple-reg:
Equation (daily):Number of patients = 61.7 + 12.67*RelHumi – 5.1*MAX-temp
![Page 39: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/39.jpg)
multiple-reg:
Equation (daily):Number of patients = 61.7 + 12.67*RelHumi – 5.1*MAX-temp
IMPORTANT: This equation works only with z-scores of independent variables…
![Page 40: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/40.jpg)
multiple-reg:
Equation (daily):Number of patients = 61.7 + 12.67*RelHumi – 5.1*MAX-temp
IMPORTANT: This equation works only with z-scores of independent variables…
Prediction example…predict the number of patients in a day where the maximum temperature was 25 Celsius and we had 50% relative humidity…
Z-score transformation = (x – μ) / σ
𝑍𝑚𝑎𝑥𝑇𝑒𝑚 =𝟐𝟓 − 21.57
18.88= 0.79
𝑍𝑟𝑒𝑣𝐻𝑢𝑚 =𝟓𝟎 − 53.29
74.37= −0.4
Number of patients = 61.7 + 12.67*(-0.4) – 5.1*(0.79) = 52.6 ≈ 53 patients
![Page 41: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/41.jpg)
Time series analysis…
![Page 42: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/42.jpg)
Type of preditions
Predictions
Nu
mb
er o
f pa
tients
…relationship respect to other variables
…the historical data of the variable
![Page 43: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/43.jpg)
Time series:
![Page 44: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/44.jpg)
Time series:
![Page 45: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/45.jpg)
Time series:
Type of Series
Stationary
Non-stationary
![Page 46: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/46.jpg)
Time series:
Constant trend
![Page 47: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/47.jpg)
Time series:
homogeneousvariance
![Page 48: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/48.jpg)
Time series:
Autocovariance
![Page 49: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/49.jpg)
Time series:
Methods for non-stationary time series:
❑Moving Average (MA)
❑Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
The more useful for realistic datasets
![Page 50: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/50.jpg)
Time series:
![Page 51: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/51.jpg)
Time series:
Autocovariance
Clearly a non-stationary series…
![Page 52: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/52.jpg)
Time series:
![Page 53: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/53.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
![Page 54: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/54.jpg)
Time series:
❑ Moving Average (MA)
User-defined parameter(s): n
![Page 55: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/55.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
Two(2) moving periods…
Forecasting one next period only
Range of observed variables, syringe consumption, over time…
Print the results starting from the indicated cell…
![Page 56: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/56.jpg)
Time series:
![Page 57: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/57.jpg)
Time series: The algorithm performance indicators…
Methods for non-stationary time series:
❑ Moving Average (MA)❑ Errors…
𝑴𝑨𝑬 =𝟏
𝒏
𝒕=𝟏
𝒏
𝑨𝒕 − 𝑭𝒕Mean Absolute Error(MAE)
Mean Squared Error(MSE) 𝑴𝑺𝑬 =𝟏
𝒏
𝒕=𝟏
𝒏
𝑨𝒕 − 𝑭𝒕𝟐
Optimum
≈ 𝟎
Mean Absolute Percentage
Error(MAPE)
𝑴𝑺𝑬 =𝟏𝟎𝟎
𝒏
𝒊=𝟏
𝒏𝑨𝒕 − 𝑭𝒕𝑨𝒕
≈ 𝟎
𝒆𝒕 = 𝑨𝒕 − 𝑭𝒕
≈ 𝟎
Error = 27 – 24 = 3
![Page 58: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/58.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)…3 moving periods…
MA-3 performance is worst compared to MA-2…
Learning: One can make more accurate predictions using MA-2
![Page 59: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/59.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
![Page 60: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/60.jpg)
Time series:
The weights should be defined in a column…
Two(2) moving periods
![Page 61: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/61.jpg)
Time series:
![Page 62: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/62.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
![Page 63: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/63.jpg)
Time series:
How do we optimize this user-defined parameter in order to reach the best possible MAE?
How do we optimize this user-defined parameter in order to reach the best possible MAE?
Let alpha as “zero”…it will be optimize
![Page 64: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/64.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Best MAE so far…!
![Page 65: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/65.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Real Stat assumes:
b1 = 0F1 = emptyBetter results…
Two(2) user-defined parameters:Alpha and Beta
![Page 66: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/66.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Let it as zeroLet it as zero
Optimizing the MAE…
![Page 67: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/67.jpg)
Time series:
Very similar to the SES…due to there is no trend (+ or -) within the dataset…DES would performs better when data exhibits certain trend…
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
![Page 68: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/68.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Three parameters…alpha, beta and gamma
Seasonality is new
Seasonality: it is traditionally defined by the number of observations within a year…
![Page 69: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/69.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
There are 12 observations every year
![Page 70: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/70.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
12 observations during each revised year…
![Page 71: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/71.jpg)
Time series: Forecasting start right after the first “s” cycle (13th
observation)
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Better that SES and DES
![Page 72: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/72.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
The forecasting equation is formed by the multiplication of the series components…
![Page 73: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/73.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
Best performance reached
![Page 74: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/74.jpg)
Time series:
Methods for non-stationary time series:
❑ Moving Average (MA)
❑ Weighted Moving Average (WMA)
❑ Simple Exponential Smoothing (SES)
❑ Double Exponential Smoothing (DES) (Holt´s Method)
❑ Holt Winter´s Additive (HWA)
❑ Holt Winter´s Multiplicative (HWM)
![Page 75: Forecasting in HCSC](https://reader034.fdocuments.net/reader034/viewer/2022042421/6260ebf870228d15ad54c16f/html5/thumbnails/75.jpg)
Thanks...
Erasmus+ KA2 Strategic Partnership2017-1-1FI01-KA203-034721Healthcare Logistics Education and Learning Pahtway