Procedure to improve the accuracy of dental implant failures by data science techniques

NANCY B. GANZ; ALICIA E. ARES; HORACIO D. KUNA

Journal of Computer Science and Technology

Facultad de Informática, Universidad Nacional de La Plata

Lugar: La Plata; Año: 2021 vol. 21 p. 146 - 156

1666-6046

Nowadays, the prediction about dental implantfailure is determined through clinical andradiological evaluation. For this reason, predictionsare highly dependent on the Implantologists?experience. In addition, it is extremely crucial todetect in time if a dental implant is going to fail, dueto time, cost, trauma to the patient, postoperativeproblems, among others. This paper proposes aprocedure using multiple feature selection methodsand classification algorithms to improve theaccuracy of dental implant failures in the province ofMisiones, Argentina, validated by human experts.The experimentation is performed with two datasets, a set of dental implants made for the case studyand an artificially generated set. The proposedapproach allows to know the most relevant featuresand improve the accuracy in the classification of thetarget class (dental implant failure), to avoid biasingthe decision making based on the application andresults of individual methods. The proposedapproach achieves an accuracy of 79% of failures,while individual classifiers achieve a maximum of72%.