A time series is the continuous data points that have occurred at some point in time. This area in machine learning is often ignored. There are several key areas in the field of machine learning where time plays an important part in accordance to the data.
so, As all the statistics data is concern the time series data is collected from the real life thing that we are interest in at the particular time.
What is Time Series?
Time series can be characterize as the succession of metrics record over a time period. Depending on the frequency a time series can be yearly, quarterly, monthly.
There are 2 components of the time series that makes it differs from the general regression problem :
- Time-dependent
- Seasonality trends
Time series forecasting can be divide into 2 methods :
- Univariate Time-series Forecasting
There are generally two variables in that one is time and other is the parameter.
- Multivariate Time-series Forecasting
It also contains multiple variables among which one is contain as the time factor and other are the multiple parameters.
Analysis of its data
When we analyse the normal data time without time as the factor our traditional statistics methods are generally able to give us the brief about the data. but in case of time series to get the better picture of the data it require to create a model that is capture time series from which we are able to understand the underlying causes in the end.
the data helps us seek the “How and Why” reason behind the data and this is often involves creating a statistical and mathematical model which is able to provide descriptive description from the provided data
forecasting
It involves taking the model and fitting it on the historical data utilizing them to anticipate the future. however, At the initial step past observations were analyze to develop a suitable mathematical model which captures the underlying data generating process for the series. In the second step the future is anticipated utilizing the model. This methodology is especially valuable when there is an absence of an illustrative model.
Components of time series behavior
- Trend
- Seasonality
- Cycle
- Variation
- Irregularities
There can also the several irregularities or outlier with the data which can be related back to the real life data
Trend
It is the overall long term direction of the series these usually go up or down depending upon the discernible trend
Seasonality
It occurs when there is some sort of the repetitive behavior in the data which occur at regular interval of time
Cycle
These generally occur when there is an up and down trend that is not affected by seasonality. Cycle can notoriously vary in length which makes them somewhat unpredictable to detect than seasonality
Variation
In every data there will be any sort of variation either small or big sometimes there are also upwards or downwards trends.
Irregularities
There can be several reasons for this to occur in the data pattern.
Time Series forecasting applications
From the beginning till now we are able to understand how the time series model works by analysing the trend and the seasonality from the existing data and forecasting the future predictions.
Such type of forecasting can be useful in many industries such as:
- Financial : Sales forecasting, inventory analysis, stock market price predictions, price estimation according to market
- Weather : Temperature estimation, climate change, seasonal shift recognition, weather forecasting.
- Network data : Network usage prediction, anomaly or intrusion detection, predictive maintenance
- Healthcare : census prediction, insurance benefits prediction, patient monitoring.
These are some of the mainstream examples but there are several that are still not discovered for example people spending time on e-commerce websites day, night, morning.
We are only seeing the starting of the growth of such data keeping into the databases by the companies
Conclusion
Analysis of such patterns is the first step of converting non-stationary data in to stationary data , so that the statistical forecasting methods could be applied.
Written By : Mukut Khandelwal
Reviewed By: Vikas Bhardwaj
