- SVM classifier for microarray carcinoma data | MATLAB
- SVM with linear and non-linear Gaussian Kernel was built for classification of carcinoma data
- Leave one out and Leave ten out cross validation techniques was used to prevent over-fitting and under-fitting of data
2. SVM classifier for diabetes data |MATLAB
- SVM with linear, polynomial and Gaussian kernels was developed for classification and performance was compared
- Effect of different cost and gamma values on accuracy of Gaussian Kernel classification was observed
- Optimal values of gamma and cost, over a range, was determined by observing the sensitivities of the performance accuracy
- Implemented ANN for same data and compared performance between locally optimum and globally optimum solutions
3.Analysis of accuracy of classification algorithms with feature extraction and dimensionality reduction for epileptic seizure detection |MATLAB
- Built a 10-layer feed forward Artificial Neural network , Decision Tree and Support vector machine using Gaussian Kernel to classify seizure data with 11500 samples from EEG recordings
- In feature extraction method, the statistical EEG features was calculated for each EEG sub-band and used as features
- In dimensionality reduction method, principal component analysis was performed. The effect of number of principal components on classification accuracy was also studied
4.Study of the effect of flexion angle on neck musculoskeletal disorders using computational model |Opensim
- Analyzed the variation in moment generating and force generating capacities of specific neck muscle group for different movements
- Study finding revealed why 400 flexion angle is commonly referred as forward posture angle from the net fiber force plots
- Study findings confirmed that trapezius muscle undergoes maximum physical exertion during flexion from force generating capacity of muscle plots
5.Comparison of efficiency of different parameter estimation methods for a single neuron model |MATLAB
- Estimated the parameters of Izhikevich neuronal model using MATLAB’s optimization algorithms implementing quadratic program solving, pattern search and least square estimation using methods of lumped parameter estimation and non-lumped parameter estimation
6.Analysis of an ARMA dynamic system for excitation requirements and estimation accuracy in the presence of noise |MATLAB
- For a specified 3rd order ARMA system, system identification using Least Square batch algorithm, over parametrization and under parametrization was carried out
- Analyzed the system’s performance for different signals with varying excitation and noise levels