The Applicability of Regression Analysis and Artificial Neural Networks to the Prediction Process of Consistency and Compaction Properties of High Plastic Clays

Heuristics‐enhanced geospatial machine learning (SaaS) of an ancient Mediterranean environment

“…The remote sensing data models that were identified via both the machine learning and statistical modeling approaches do not support the models developed (Dastbaz et al, 2018), even though the same Landsat datasets were used in our models. Our models were less than 50% as accurate, which leads to questions concerning the significant deviation concerning similar models used by others (Arama et al, 2020;Armstrong et al, 2007;Louca et al, 2018;Qu et al, 2018). In our spatial model, the data were not significant enough to warrant the use of the model in its current form with a F I G U R E 2 Potential for ground movement (GM) in cm from the generalized additive model based on the laboratory dataset (within the GPS polygon Figure 1.)…”

“…Investigating the character of remotely sensed big‐data becomes an essential need to help verify (if available) laboratory datasets for improved model accuracy, especially where data are scarce. These are the first models to predict ground movement in soils older than 200 million yr, a first for Mediterranean climates, a first for Australia, and possibly a first for the Southern Hemisphere (Arama et al., 2020). The model developed from laboratory data is reliable for industry use concerning ground movement within the GPS coordinates of the UTM zone 50 polygon (Figure 1).…”

“…The parent material, which includes an igneous formation located west of the Darling Range dating back at least 2,600 million yr, informs the soil physical and geological landscape of that locality; broadly, framework silicate–based, iron‐rich sands that gradually increase to more calcareous mixtures near the wave‐affected coastline (McArthur, 2004). Modeling of geological age within this geographical location constitutes a broad range in terms of soil era and thus has had limited uptake concerning the building of new AI‐based models, unlike other industries where computational advancements may be rapidly outstripping concise data collection strategies but conversely are having significant positive statistical gains in the order of 5–10% (improvement in prediction accuracy) (Arama et al., 2020; Guo et al., 2021; Yu et al., 2021) (Figure 1).…”