5

Noise does not have negative connotations in statistics or machine learning. In both cases, we are dealing with random variables, so if $\mathbf{x}$ is a random variable, then any function of it $f(\mathbf{x})$ will be a random variable as well. In such a case, the output of the function would be random, non-deterministic, or “noisy”. It is not the loss ...


3

My guess is that Bayesian (Logistic) Regression is what you're looking for. In a regular Logistic Regression you find the (model) parameters $\beta$ that maximise the likelihood of your data (see Maximum Likelihood estimation). In a binary classification you use those parameters to calculate the probability $p$ of a given (test) sample $X_{test}$ of ...


2

Have you considered augmenting your accelerometer data by applying rotations so that the trained classifier is robust against changes in sensor orientation? This approach has been successfully tried in similar tasks, e.g. here One factor that can introduce label-invariant variability of wearable sensor data are differences in sensor placement between ...


2

As geedigit notes, it depends what is important to your application. I doubt recall would be a good metric to monitor because it gets the conditioning incorrect. In a probabalistic sense, recall is $$ P(\hat{y} = 1 \vert y=1) \>. $$ In essence, of all the positive outcomes, what proportion did you predict were positive? Note this quantity necessarily ...


1

A alternative solution to transforming the data inputs is to let the Gaussian Process (GP) model directly handle the categorical inputs. The model kernel can be combined as a sum of a categorical kernel and a regular kernel of the form: K((x1, c1), (x2, c2)) = K_cont_1(x1, x2) + K_cat_1(c1, c2) + K_cont_2(x1, x2) * K_cat_2(c1,c2) A readily available ...


1

should I use only the granular level detail variables in my ML model and ignore other levels of same variable? because I guess it would be correlated It depends on how much variation there is within the heirarchies. I'm not really sure what the difference is between product family and product segment is so let me give you a different example. Cars have ...


1

a) The probability values from a well-calibrated$^{\dagger}$ model will be your friend. If you have the budget to contact $500$ people, contact the $500$ people most likely to respond. Many machine learning methods output probability values that one can map to a discrete outcome by using a threshold, but doing so has major drawbacks, such as not being able ...


1

Given a city you can immediately infer both the state and a country. So the latter two features give you no additional information whatsoever! Because of that, the simple answer would be to remove state and city from the dataset. However, there are some caveats here. First of all - what's the original problem you're trying to solve? If you're interested in ...


1

I'm not sure I understood what you are trying to do, but the first thing that came to my mind was to use the attitude angles in a coordinate transformation. Unfortunately, it is probably not as simple as you describe it, as is nicely presented in this SO question. What you can do is to transform your accelerations from its body-fixed coordinate system into a ...


1

SVM's are a kernel method. A kernel can be in an infinite dimensional feature space. After evaluating data points in this "higher" dimension, we can then perform linear inference on them Mercer's Theorem. SVM's specifically use the higher dimensionality to aid in drawing a hyperplane which can separate the data Kernel trick. The bad part about this ...


1

I'm a beginner at this topic, but we dealt with it in a course of my university. In your case, you could remove near zero variance variables and aggregate certain categories. I would recommend you to read the Chapters 3.4 and 3.6 of the following book: https://bradleyboehmke.github.io/HOML/engineering.html#feature-and-target-filtering


Only top voted, non community-wiki answers of a minimum length are eligible