Guide to make the Best choice among Time Series Algorithms

1. Long Short-Term Memory (LSTM) Neural Networks: LSTMs are a type of Recurrent Neural Network (RNN) that are well-suited for time series forecasting.

2. Autoregressive Integrated Moving Average (ARIMA) and Seasonal Autoregressive Integrated Moving Average (SARIMA): ARIMA and SARIMA are traditional time series forecasting methods that have been widely used for many years and are still popular today.

3. Prophet: Prophet is a forecasting model developed by Facebook that uses Bayesian techniques to model time series data with multiple trend and seasonality components.

4. Gradient Boosting Machines (GBM): GBMs are a type of decision tree-based ensemble method that can be used for time series forecasting by modeling the relationship between past observations and future outcomes.

5. Support Vector Regression (SVR): SVR is a type of regression algorithm that can be used for time series forecasting by modeling the relationship between past observations and future outcomes.

6. XGBoost: XGBoost is a type of gradient boosting algorithm that is well-suited for time series forecasting and has been widely used in many competitions and real-world applications.

Ultimately, the choice of algorithm will depend on the complexity of the data, the resources available for model training and prediction, and the desired accuracy and interpretability of the model.

In many of the cases in time series forecasting requires retain of information to give accurate and precise prediction in such cases algorithms in such cases where the retention of past data is required if algorithms like ARIMA/SARIMA is used the accuracy and precision will take a toll as it will prediction neglecting the past data which will eventually lead to loss of characteristic and variability of data. ARIMA (AutoRegressive Integrated Moving Average) and SARIMA (Seasonal AutoRegressive Integrated Moving Average) are traditional time series forecasting methods that have been widely used for many years. However, in recent years, with the advancement of deep learning models, such as LSTMs, these traditional methods may sometimes be less accurate compared to more advanced techniques.

The accuracy of ARIMA and SARIMA depends on the quality of the data and the skill of the modeler in selecting the right parameters and order of differencing. If the time series data is relatively simple, with clear patterns and trend, these traditional methods may still provide accurate results. However, if the data is more complex, with multiple underlying patterns and noise, LSTMs and other deep learning models may provide more accurate predictions.

It is worth noting that ARIMA and SARIMA have the advantage of being easier to interpret and implement compared to deep learning models, and they can still provide good results for certain time series forecasting problems. Ultimately, the choice of model will depend on the complexity of the data and the resources available for model training and prediction.