Machine Learning for Healthcare Applications. Группа авторов
“Probabilistic Neural Network” classification with 96% accuracy.
They used the scenario of predicting the consumer’s choice based on EEG signal analysis [13] while viewing the trailers which resulted in finding significant gamma and beta high frequencies with high correlation to participants and average preferences.
Participants were assessed on self-arousal and valence features while watching particular scenes in a movie [14]. They analyzed the data while factoring in 5 peripheral physiological signals relating them to movie’s content-based features which inferred that they can be used to categorize and rank the videos.
Here 19 participants were shown 2 colors for an interval of 1 s and during the time EEG oscillations were analyzed [15] on Neural mechanisms for correlations of color preferences.
They had 18 participants who were subjected to a set of choices and analyzed their Neuro-activity and Eye-tracking activity to brain-map regions associated with decision making and inter-dependence of regions for the said task [16]. They concluded with high synchronization amongst frontal lobe and occipital lobe giving major frequencies in theta, alpha and beta waves.
They are trying to establish a bond between Neuro-signals and the learning capacity of a model software [17] while assuming that the model has the capability to train itself for dominant alpha wave participants.
“Independent Component Analysis (ICA)” to separate multivariate signals coming from 120 channels of electro-cortical activity [18]. This was done to convert those signals into additive subcomponents. Patterns of sensory impulses were recorded which matched movement of the body.
They have used filter is as a stabilizing and filtering element in the ECG data of 26 volunteers and then applied Approximate Entropy on it for inter-subject evaluation of data as the part of a retrospective approach [19] while adding truthfulness to Entropy windows for its stable distribution. This filter is very extensively being used in Signal processing which led us to adopt it.
The study [20] is an experiment on ECG signals of 26 participants where approximate entropy method is implemented for examining the concentration. Approximation entropy window was taken less for intra-patient comparing to inter-patient and for filtering the noisy signals S-Golay method was implemented.
They have innovatively preprocessed the ECG signal using S–Golay filter technique [21]. With both quadratic degrees of smoothing and differentiation filter methods combinedly has processed ECG signals having sampling rate 500 Hz with seventeen points length.
A very unique “double-class motor imaginary Brain Computer Interface” was implemented with Recurrent Quantum Neural Network model for filtering EEG signals [22].
In the paper [23] using the S-Golay filter, the artifacts due to blinking of eyes are found out and it is eliminated adapting a noise removal method.
3.3 Methodology
3.3.1 Bagging Decision Tree Classifier
Among the many Machine Learning algorithms, this method forms a group of algorithms where several instances are created of black-box estimators on variable subsets from the base training set after which we aggregate their solo predictions to form a resultant prediction. This process is used as a path to minimize the variance of the foundation estimator i.e. a decision tree by including randomization within its creation process and building an ensemble from it. In multiple scenarios, this method consists a simple path to improve with regard to a single model, avoids making it a necessity to acclimatize to a foundation algorithm. It works best with fully developed decision trees as it reduces overfitting in comparison to boosting methods which generally work best in shallow decision trees. This classifier comes in many flavors but majorly differ from each other by the path that they draw variable subsets of the training set. In our case samples were extracted with replacement called as Bagging.
3.3.2 Gaussian Naïve Bayes Classifier
This classifier is based on probability which is combined within a Machine Learning model. Hence, it is based on “Bayes Theorem” which states that, we can derive the probability of an event1 given that a retrospective event2 has happened. Here, event2 is the witness and event1 is the hypothesis. The assumption here is that the features are non-dependent which means that the existence of one feature does not affect the other which is why it’s called Naïve. When predictions allocate a continuous value without being discrete, we can ascertain that those values are derived from gaussian distribution. Following is the general formula for Bayes theorem (3.1).
Since our case has a different set or input, our formula for this implementation changes to Equation (3.2).
3.3.3 Kernel Support Vector Machine (Sigmoid)
The separable data with non-linear attributes cannot be tackled by a simple Support Vector Machine algorithm due to which we use a modified version of it called Kernel Support Vector Machine. Essentially in K-SVM it presents the data from a non-linear lower dimension to a linear higher dimension form as such that the attributes belonging to variable classes are assigned to different dimensions. We use a simple Python-SciKit Learn Library to implement and use K-SVM.
For training purposes, we use the SVC class of the library. The difference is in the values for the Kernel parameters of SVC class. In simple SVM’s we use “Linear” for Kernel parameters but in K-SVM we use Gaussian, Sigmoid, Polynomial, etc. wherein we have used Sigmoid.
The only limitation observed in our case is that though this method achieves the highest accuracy but not up to the mark. Hence more advanced models like Deep Learning may be applied in near future for more concrete results.
3.3.4 Random Decision Forest Classifier
It is a variant of supervised machine learning algorithm founded on the schematic of ensembled learning. Ensemble learning is an algorithm where you join multiple or single algorithm into multiple types of algorithms of multiple or same variant to create a complex and advanced prediction model. It also combines many algorithms of same variant as decision trees, forest trees, etc. so the name “Random Forest”. It is used for regression and classification tasks.
The way it works is it picks a part of the dataset and builds a decision tree on these records, and after selection of number of trees you want this process is repeated. Each tree represents the prediction in that category for which the new record belongs. The only limitation here is that there forte lies in their complexity and for that we need substantial computing resources when huge number of decision trees can be brought together which in turn will better train themselves.
3.4 System Setup & Design
We have used an Emotiv EPOC+ biosensor device for capturing Neuro-Signals in the following manner. Figure 3.1 represents the channels on the brain from signals collected and the equipment used for collection. The signals are collected from 14 electrodes positioned at “AF3, AF4, F3, F4, F7, F8, FC5, FC6, O1, O2, P7, P8 T7 and T8” according to International 10–20 system viewed in the figure below. There are reference electrodes positioned above ears at CMS and DRL. By default, the device has a sampling frequency of 2,048 Hz which we have down-sampled to 128 Hz per channel. The data acquired is transmitted using Bluetooth connectivity to a system. Before every sample collection