Neural network-based system for optimizing process parameters of semiconductor compounds

Finding the optimal parameters in anionic co-doped titanium dioxide (TiO₂) is an important task in the compound preparation on either photocatalytic-oriented or mechanical-preferred properties. This work proposes a neural network-based system to optimize the process parameters of the deposition of TiC_xO_yN_z films. The proposed system comprises three stages, which are data processing, parameter training and rule mining. The data processing phase classifies instances of sputtering gases into groups. The parameter training phase optimizes the parameters of the deposition of TiC_xO_yN_z. Three layers - input, hybrid hidden and output layers - reside in this phase. Both feed-forward and feed-back connections between the input layer and the hidden layer and between the hidden layer and the output layer are established. The mining phase uses the C4.5 algorithm to construct a decision tree and develop rules. An illustrative example demonstrates the performance of the proposed system. The results indicate that the proposed neural network system exhibits superior performance to optimize the deposition parameters of TiC_xO_yN_z and mine rules for so doing.