Adding Time Series Lime Explainer

[gganapavarapu]

This PR is to add Time Series Lime (TSLime) Explainer algorithm for forecasting, time series classification and regression use cases.

Short description of the algorithm is: Advanced time series models are complex and often hard to interpret. Time Series Local Interpretable Model-agnostic Explainer (TSLime) is a model-agnostic local time series explainer. It explains a time series model behavior using an approximate linear model as surrogate. TSLime approximates model response by evaluating the model over time series samples generated by applying time series perturbation techniques. The explanation produced by TSLime is the weights of the linear model over different time point observations. The relative signed value of the explanation is indicative of model sensitivity at temporal resolution. TSLime uses the recent time context length for the local surrogate model fitting.

Required Dependencies:

• numpy

• pandas <= 1.4.3

• scikit-learn

• scipy

• algorithm folder is added here

• example notebooks are added here

• updated the examples readme here with univariate time series classification and multivariate forecasting notebook details.

• This explainer's dependencies are subset of TSICE. So, included it in TSICE job here.

[gganapavarapu]

rbm units were failing with latest numpy (1.25.x). Fixed the version to 1.24.x for rbm algorithm. The issue was not from current TSLime algorithm changes.

[vijay-arya]

Thanks @gganapavarapu