Notebook Difficulty Level: ★★★★☆
“Mathematics is everywhere!” 😏
Despite not being the most original sentence, this idea is almost a universal truth, something that makes it the ideal candidate to start our introductory text.
Probability and Statistics is one prominent branch of (Applied) Mathematics identifiable in a wide range of segments of our society, ranging from the daily evaluation of future meteorological conditions to epidemiological studies while evaluating the risk that a subject presents of contracting a disease.
With such a diversified set of possibilities and exciting opportunities, Probability and Statistics also provide extremely important tools to computational sciences, contributing for the creation of a notable group/family of Machine-learning algorithms based on Bayesian/probabilistic inference .
One of the simpler members of this family is the Naive Bayes classifier, which belongs to the group of supervised machine-learning algorithms supported by theBayes Theorem , i.e., in the “assumption of conditional independence between every pair of features given the value of the class variable” ( further details in scikit-learn official page ).
The “naive” term suits extremely well, considering that the Bayesian assumption, regarding the conditional independence between features, is commonly not true.
This Jupyter Notebook will be dedicated to present a practical application of Bayesian/probabilistic inference through the training and evaluation of a Naive Bayes classifier focused on the detection of Fist activity .
As demonstrated in other Jupyter Notebooks belonging to Train and Classify category ( Signal Classifier – Distinguish between EMG and ECG and Rock, Paper or Scissor Game – Train and Classify ), machine-learning algorithms can be used to distinguish signals or identify movements, now it will be explored what they can offer while facing an event detection challenge.