justcodebase
commented
4 months ago Ah. I just took a look at your profile. It seems you are representing The Financial Engineering Department of The AmirKabir University of Technology (Polytechnic of Tehran, Iran). I was not aware about this fact. I have to admit that I have no idea about any potential constraints (and if such constraints exist at all) that may relate to this fact. I am very sorry. I will try to learn more about it. I am just a private individual interested in financial subjects. Again, I will try my best to learn more about the current environment. I really mean it. I am not looking for any troubles. I am very sorry again. Just wanted to say that your public repos look very interesting. I wish you best.
shayandavoodii
Would you have any suggestion regarding a more efficient clustering method?
I asked Mistral AI and was informed about:
Distance-based methods: These methods measure the similarity between time series using a distance metric, such as Euclidean distance, Dynamic Time Warping (DTW), or Longest Common Subsequence (LCSS). Once the distances between all pairs of time series have been computed, a clustering algorithm such as K-medoids or hierarchical clustering can be applied.
Model-based methods: These methods assume that each time series is generated by a probabilistic model, such as a Gaussian Mixture Model (GMM) or a Hidden Markov Model (HMM). The parameters of the model are estimated from the data using techniques such as Expectation-Maximization (EM), and the time series are assigned to clusters based on their probability of being generated by each model.
Shape-based methods: These methods focus on the shape of the time series, rather than their absolute values. For example, the time series may be normalized to have zero mean and unit variance, and then clustered based on features such as their Fourier coefficients or wavelet transforms.
Subspace clustering methods: These methods assume that the time series lie in a low-dimensional subspace of the high-dimensional space in which they are embedded. The subspace is estimated using techniques such as Principal Component Analysis (PCA) or Singular Value Decomposition (SVD), and the time series are clustered based on their proximity in the subspace.
Recently I have been reading about HiddenMarkovModels.jl (https://github.com/gdalle/HiddenMarkovModels.jl) by @gdalle and a few years ago (under a different name on Julia Discourse) I was in touch with @MommaWatasu, who is the author of some algorithms still quite unique in the Julia Ecosystem (https://docs.juliahub.com/General/HorseML/stable/Tutorial/Clustering/#).
Do you have maybe opinion about those methods?