Zomato Interview Questions | Seasonality
Question
How can you deal with different types of seasonality in time series modeling? Also, mention a few examples to support your answer.
in progress
0
Statistics
4 years
1 Answer
823 views
Great Grand Master 0
Answer ( 1 )
Seasonality – recurring but not necessarily periodic data patterns, capturing true seasonality greatly enhances model accuracy,
We can deal seasonality in majorly two ways:
1. With simple statistical methods
2. Tuning based on modeling results
The statistical method involves:
a. De-trending data with a centered moving average the size of the estimated seasonality.
b. Isolating the seasonal component with one moving average per relevant time-step
Tuning based on modeling involves:
a. Seasonal-ARIMA models
b. Fourier series model where seasonality based of linear combination of sine and cosine terms
c. Exponential Smoothing models