The choice of a machine learning model depend upon the use case and resources.In general most widely used models for time series analysis are:

LSTM is a type of RNN that is able to capture long-term dependencies in time series data.It does this using "memory cells".

Long Short-Term Memory (LSTM)

GRU is another type of RNN that is similar to LSTM, but it has fewer parameters and is generally easier to train

Gated Recurrent Unit (GRU)

SRU is a simplified version of LSTM and GRU that has fewer parameters.It has been shown great performance on a wide range of time series tasks.

Simple Recurrent Unit (SRU)

Echo State Network is a type of RNN that is easy to train and has a simple structure. It is particularly well-suited for tasks with large, noisy time series data.

Echo State Network (ESN)

It can capture long-term dependencies using dilated convolutions, which allow the network to "look back" in time over a larger context window.

Temporal Convolutional Network (TCN)