Global Journal of Human-Social Science, B: Geography, Environmental Science and Disaster Management, Volume 22 Issue 3

structure is obtained, evidencing the association that the algorithm makes between the many variables. Still, it was possible to notice that the mechanical characteristics of soil compaction ( ρ dmax and W o ) were decisive for the initial division of the groups, as well as the granulometry (mainly the percentage of silt). The chemical classification of aluminum saturation (S, %) indicated by Prezotti (2013) also showed a relationship with the clustering of the soils, indicating that the clustering considered, in fact, different types of characteristics of fine-grained tropical soils (physical, mechanical and chemical). It was also noted that soils with mechanical clayey behavior (higher optimum moisture and lower dry apparent density) showed a more homogeneous behavior forming a cluster composed entirely of soils of the ML class (from the USCS classification, which is based on the LL and the PI) and with very similar characteristics of granulometry (percentage of silt and fine sand), in addition to the geotechnical compaction characteristics indicated. In the group formed by soils of sandy behavior (B), the subdivision was done also by considering, in addition to the parameters considered in group A, the organic matter content, and it was noticed that the association between several other variables was also considered, since there was a tendency towards lower or higher values in some parameters, for example: in terms of total permanent deformation ( ε p ) in subgroup - B-2-a: soils with ε p larger (1.63 – 7.19 mm) than those of group B-2-b ( ε p between 0.78 – 3.37 mm) were included; the suction in the first at the air intake point associated with macropores ( ψ b1 ) tended to be slightly higher in the soils of group B-2-a (3.4 – 4.8 kPa) when compared to B-2-b (1.2 – 3.8 kPa). The recognition of the similarity between some pairs of soils proved the validity of the hierarchical clustering technique since several variables with similar values were identified between them, even though they were of a different nature (physical, chemical and mechanical), the most recurrent being: LL, PI, %Clay, % silt, c’, e’, pH KCl , δ , S b , Ψ b1 , Ψ res1 , S res1 , W o , K2TC, K2TD, ε p . The comparison of images applying the techniques of data science also corroborated, since satisfactory results were obtained, congruent with those obtained in cluster analysis, with rare exceptions. It is assumed, therefore, that there is correspondence between the mineralogical characteristics visualized in the images (iron oxides, rock fragments, quartz grains, etc.) with the results of geotechnical tests. Finally, it is concluded that the application of cluster analysis by hierarchical method, as well as the comparison of microscopic images, using the tools of Data Science, showed useful techniques and tools for the cluster analysis of fine-grained tropical soils since it portrayed the similarity of behavior of different soils considering several geotechnical aspects. A CKNOWLEDGEMENTS The authors thank the Foundation for Science and Technology of Pernambuco (FACEPE) for the financial support (doctoral scholarship to the first author) and incentive to carry out this research. The research was carried out through the INCT REAGEO Project through the partnership signed between GEGEP (Geotechnical Engineering Group of Disasters and Plains) of THE PPGEC of UFPE (Federal University of Pernambuco) and the Geotechnics Laboratory of COPPE/UFRJ (Federal University of Rio de Janeiro). R eferences R éférences R eferencias 1. Boszczowski, R. B., & Ligocki, L. P. (2012). Chapter 8: Características Geotécnicas dos solos residuais de Curitiba e RMC. In Twin Cities – Solos das Regiões Metropolitanas de São Paulo e Curitiba. D’Livros. ISBN: 978-85-86438-40-0. [In Portuguese]. 2. Camapum de Carvalho, J., de Rezende, L. R., Cardoso, F. B. F., Lucena, L. C. F. L., Guimarães, R. C., & Valencia, Y. G. (2015). Tropical soils for highway construction: Peculiarities and considerations. Transportation Geotechnics, 5, 3– 19 . https://doi.org/10.1016/j.trgeo.2015.10.004. 3. Chen, C. Härdle, W. K. Unwin, A. 2007. Handbook of Data Visualization. Springer-Verlag Berlin Heidelberg. 936 p. DOI: 10.1007/978-3-540-33037 -0. 4. Chesworth, Ward et al. 2008. “Tropical soils”. Encyclopedia of Soil Science , [s.l.], p.793-803. Springer Netherlands . http://dx.doi.org/10.1007/978- 1-4020-3995-9_607. 5. Coutinho, R. Q., & Sousa, M. A. S. (2021). Analysis of the applicability of USCS, TRB and MCT classification systems to the tropical soils of pernambuco, Brazil, for use in road paving. In Advances in transportation geotechnics IV. Lecture Notes in Civil Engineering, Vol. 164. Springer. https://doi.org/10.1007/978-3-030-77230-7_29. 6. Coutinho, R. Q., Silva, M. M., Santos, A. N. D., & Lacerda, W. A. (2019). Geotechnical characteriza- tion and failure mechanism of landslide in granite residual soil. Journal of Geotechnical and Geoenvironmental Engineering, 145(8), 05019004. https://doi.org/10.1061/ (ASCE)GT.1943-5606.0002 052. 7. CPRM, Companhia de Pesquisa de Recursos Minerais. 2014. Geodiversidade do estado de Pernambuco . ISBN 978-85-7499-141-2. Recife. [In Portuguese]. 8. Dalla Roza, A. E. Motta, L. M. G. Classificação MCT com relação ao comportamento resiliente e deformação permanente em solos do Mato Grosso. © 2022 Global Journals Volume XXII Issue III Version I 23 ( ) Global Journal of Human Social Science - Year 2022 B Clustering of Fine-Grained Tropical Soils using Data Science Tools Applied to their Geotechnical Properties