Gints Jekabsons' webpage

Regression and metamodelling datasets

Here are some regression / metamodeling datasets (including some of those used in our publications). All the datasets can be also used for evaluations and comparisons of different regression modelling methods.

Datasets from Finite Element simulations

• Finite Element simulation results of four-stiffener (Design1) and five-stiffener (Design2) carbon fibre reinforced plastic (CFRP) panels and full scale fuselage structures under buckling load (4 input variables, 32 output variables). The datasets were used for building different metamodels in the WCSMO2009 paper - download.
• Finite Element simulation results of full plywood panels versus I-core plywood sandwiches under bending load (5 input variables, 3 output variables). The datasets were used for building different metamodels in the SSTA2009 paper - download.
• Finite Element simulation results of "Three-point bending load" and "Uniformly distributed load" on glass fibre reinforced plastic (GFRP) deck panels (6 input variables, 10 output variables). The datasets were used for building different metamodels in the ASMO-UK/ISSMO2008 paper - download.
• Finite Element simulation results of I-core, OS-core, V-core, Z-core, OC-core, C-core all-metal laser welded sandwich panels under bending load (6 input variables, 24 output variables). The datasets were used for building different metamodels in the DFE2008 paper and Machine Learning book chapter - download.
• Finite Element simulation results of glass fibre reinforced plastic (GFRP) columns under buckling load (4 input variables, 1 output variable). The dataset was used for building different metamodels. The archive includes also the results of the modelling methods - download.

The Finite Elements simulation data sets were generated at the Institute of Materials and Structures.

Other datasets

• Dataset of Wi-Fi Received Signal Strength measurements for indoor positioning via location fingerprinting techniques - download at the Wi-Fi based positioning page.
• Dataset of weather in Riga, Latvia in 2007. The dataset was used for building regression models predicting mean temperature from 11 other variables - download.

Many different data sets can be found in the following repositories: UCI, Torgo, Delve, StatLib, BU.