• Exponential Regression - WMAPE, Explained Variance, Max_Error, Mean Squared Log Error
• Generalized Linear Models - bic
• Multiple Linear Regression - R^2, MSE, RMSE, MAE, MAPE, WMAPE, Explained Variance, Max_Error
• Multilayer Perceptron - WMAPE, Explained Variance, Max_Error
• Support Vector Machine - R^2, MSE, RMSE, MAE, MAPE, WMAPE, Explained Variance, Max_Error
• Bi-Variate Geometric Regression - WMAPE, Explained Variance, Max_Error
• Bi-Variate Natural Logarithmic Regression - WMAPE, Explained Variance, Max_Error
• Polynomial Regression - WMAPE, Explained Variance, Max_Error
Unied regression also supports the application function library (AFL) state.
Unied Regression
• Unied Exponential Smoothing (new)
Unied Exponential Smoothing provides a unied input for the following functions, also massive call model
is enabled inside the function.
• SESM (Single Exponential Smoothing)
• DESM (Double Exponential Smoothing)
• TESM (Triple Exponential Smoothing)
• BESM (Brown Exponential Smoothing)
• AESM (Auto Exponential Smoothing)
• MSESM (Massive Single Exponential Smoothing)
• MDESM (Massive Double Exponential Smoothing)
• MTESM (Massive Triple Exponential Smoothing)
• MBESM (Massive Brwon Exponential Smoothing)
• MAESM (Massive Auto Exponential Smoothing)
Unied Exponential Smoothing
• VARMA (new)
The vector autoregressive moving average models (VARMA) are the vector form of autoregressive
integrated moving average (ARIMA). VARMA is used to examine the relationships among several variables
in multivariate time series analysis, while ARIMA is used in univariate time series.
VARMA
• Discrete Wavelet Transform (new)
Discrete wavelet transform (DWT) was introduced in the late 1980s. DWT transform decomposes a given
signal into a number of sets, including an approximation coecient set (cA) and multiple detail coecient
sets (cD). These coecient sets are obtained by convolving the signal with the low-pass lter and the
high-pass lter over time.
Discrete Wavelet Transform
• Discrete Wavelet Packet Transform (new)
Discrete wavelet packet transform (DWPT) decomposes both the approximation part and the deatil part,
while discrete wavelet transform (DWT) only decomposes the approximation part.
Discrete Wavelet Packet Transform
• Online Multi-Class Logistic Regression (new)
Online version of Multi-Class Logistic Regression. It is used when the training data are obtained for multiple
rounds and the model will be continuously updated.
Multi-Class Logistic Regression
• Dynamic time warping (DTW) (new)
88
PUBLIC
What's New in the SAP HANA Platform 2.0
SAP HANA Platform 2.0 SPS 06 Features