China
Africa
Explore chapters and articles related to this topic
Uncertainty Quantification for Skewed Laminated Soft-core Sandwich Panels
Published in Sudip Dey, Tanmoy Mukhopadhyay, Sondipon Adhikari, Uncertainty Quantification in Laminated Composites, 2018
Sudip Dey, Tanmoy Mukhopadhyay, Sondipon Adhikari
MARS (Friedman 1991) provides an efficient mathematical relationship between input parameters and output feature of interest for a system under investigation based on few algorithmically chosen samples. MARS is a nonparametric regression procedure that makes no assumption about the underlying functional relationship between the dependent and independent variables. MARS algorithm adaptively selects a set of basis functions for approximating the response function through a forward and backward iterative approach. The MARS model can be expressed as () Y=∑k=1nαkHkf(xi)withHkf(x1,x2,x3…..xn)=1,fork=1
Two New Nonparametric Models for Biological Networks
Published in K Hemachandran, Shubham Tayal, Preetha Mary George, Parveen Singla, Utku Kose, Bayesian Reasoning and Gaussian Processes for Machine Learning Applications, 2022
Deniz Seçilmiş, Melih Ağraz, Vilda Purutçuoğlu
MARS [36] is a regression model used to identify linear and nonlinear effects as well as interactions between covariates by means of piece-wise nonlinear models. In MARS, the nonlinear relation between the response variable y and the predictor x’s is described as y=β0+∑m=1MβmHm(x)+ε
Forecasting Domestic Energy Consumption for the Nordic Countries
Published in Stephen A. Roosa, International Solutions to Sustainable Energy, Policies and Applications, 2020
Samad Ranjbar Ardakani, Seyed Mohsen Hossein, Alireza Aslani
The nonparametric regression technique, MARS, is a form of the stepwise linear regression developed by Jerome Friedman in 1991 [75]. The model is much simpler in comparison to other approaches such as neural network and random forest. However, the MARS was derived from linear regression methodology; it can organize nonlinear links between the target and predictors. This methodology has been used by other researchers to predict building energy efficiency.
Prediction of current-induced scour depth around pile groups using MARS, CART, and ANN approaches
Published in Marine Georesources & Geotechnology, 2021
Mehrshad Samadi, Mohammad Hadi Afshar, Ebrahim Jabbari, Hamed Sarkardeh
Mentioned studies have shown that data-driven methods performed better than empirical and traditional approaches for prediction of scour depth around pile groups. Therefore, this study investigates the performance of MARS and CART (classification and regression tree) as white-box data-driven methods and ANN as a black-box data-driven method for prediction of current-induced scour depth around pile groups in clear water conditions. MARS is a new data-driven method that provides an explicit practical equation with a combination of piecewise linear regression functions over independent variables to predict the dependent variable. CART is a well-known DT that divides the input space of a given problem based on several deterministic conditions into several sub-spaces in the form of if-then rules. If-then rules created by CART over sub-spaces improve the accuracy of the prediction of the dependent variable. On the other hand, ANN is a popular black-box method that uses a complex network of neurons (nodes) to predict dependent variables based on independent variables.
Reliability assessment of compressive and splitting tensile strength prediction of roller compacted concrete pavement: introducing MARS-GOA-MCS
Published in International Journal of Pavement Engineering, 2022
Guodao Zhang, Naser Safaeian Hamzehkolaei, Hamed Rashnoozadeh, Shahab S. Band, Amir Mosavi
The CS and Splitting Tensile Strength (STS) of RCCP are considered the non-linear characteristics of its ingredients because they are realised only by its discrete results. In this case, selecting the most suitable forms, the model’s degree, and appropriate values of the model parameters considerably affect the efficiency and accuracy of the developed prediction model, thereby providing an effective representation of different RCCP mix parameters. Although classical DDMs provide estimation models that may have accuracy and robustness for the prediction of the mechanical characteristics of concrete, they are complicated and may have a high-cost training process. MARS is a non-parametric regression technique that can automatically model non-linearities and interactions between variables, even if no prior information on the form of the response function and/or no explicit relation between independent and dependent variables is available (Zhang and Goh 2016). Nevertheless, similar to ANNs (Momeni et al.2015), MARS implementation may have some disadvantages, i.e. slow rate of training and getting trapped in local minima. Besides, effective tuning of the model input parameters such as the maximum number of basis functions (), maximum number of interactions (), and the penalty parameter () is also a challenging task that can affect the efficiency and reliability of the MARS-based developed models. To overcome these disadvantages, the use of powerful optimisation algorithms to enhance MARS performance is of great interest. Integrated models based on the combination of DDMs and global optimisation algorithms have an attractive viewpoint of limited control factors for the implementation of novel methods and reduction of the model’s overfitting (Nieto et al.2015).
Statistical Approaches to Forecasting Domestic Energy Consumption and Assessing Determinants: The Case of Nordic Countries
Published in Strategic Planning for Energy and the Environment, 2018
Samad Ranjbar Ardakani, Seyed Mohsen Hossein, Alireza Aslani
The nonparametric regression technique, MARS, is a form of the stepwise linear regression developed by Jerome Friedman in 1991 [75]. The model is much simpler in comparison to other approaches such as neural network and random forest. However, the MARS was derived from linear regression methodology; it can organize nonlinear links between the target and predictors. This methodology has been used by other researchers to predict building energy efficiency.