Properties Prediction Of ABC₂Semiconductor Using Data Mining And Development Of Data Mining Software

Data mining is a powerful technology to find the unknown, hidden and interestingknowledge from the massive data. It has been recognized as a key research topic indatabase and machine learning. It has also aroused wide interest of scientific andindustrialcircle for its large potential application.It is an important and basic work to develop powerful software of dataming for easyusage. By using VB.NET and C#language, a datamining software is developed forthedata alaysis, data pretrement and different kinds of machine learning algorithms. Thesoftware can be easily used in the process of datamining.Most of ABC2compounds are semiconductors. They are widely developed asnonlinear optical devices. In this work, Genetic Algorithm and Support VectorClassfication were used to construct the model for prediction of the formation of ABC2system. The accuracy of cross-validation and test set are92.04%and91.67%respectively.The model for predicting the structure of ABC2is established by using k-NearestNeighbor algorithm. The accuracy of the model is92.11%. The models of prediction ofband gap and melting point of the ABC2semiconductor based on Support VectorRegression are also constructed. The RMSEs (root mean square error) of corss-validationand test set based on the models of predicting Eg are0.228and0.205respectively. TheMREs (mean relative error) of cross-validation and test set based on the models ofpredicting Tm are6.83%and5.86%respectively.