This is the age of machines and it is data that is the living soul for these machines. No machine can understand any human language, they only understand numbers and operations, numbers that have gone through complex computations.

Before understanding what Time Series Analysis is, let us first have a closer look at a Time Series Data.

**Diving into Time Series Data**

Let us understand this with an example. Let’s say India expanded its military strength of 1.5 active military personals in 2018 to 2 million in 2019. That is, of course, an increase in power, sure increase in military strength is not all to determine power but let’s suppose it is. We know that in just one-year India strengthened its military with half a million soldiers.

A closer look shows how much Indian military has strengthened in a year. Clearly, we can see something is dependent on time. And say we have reasonable records proving the change in the number over different points of time and then we call it dataset and we can use it to predict the number for future. Now, this is Time Series Analysis.

We cannot predict the result of a war with the number of soldiers but we sure can predict the number of soldiers with the data. And that is the power of data.

The simple dataset looks something like this:

**Sales Forecasting**

The graph below plots the relationship between the Date and Sales in Number variables.The Dataset shows how the Number in sales of a product changes every month

**Analysis vs Forecasting**

It is very common to see both the terms ‘Time Series Analysis ’ and ‘Time Series Forecasting’ together. What they generally mean are the 2 objectives of a Time Series Problem.

Time Series Analysis refers to the analysing of data to identify patterns and Time Series Forecasting refers to the prediction of values from the identified patterns.

**Univariate vs Multivariate Time Series Datasets**

There may not just be one factor that is dependent on time like we saw in the example of military strength where only the Number of soldiers was dependent on time or in the case of sales forecasting where only the number of sales were dependent on Time

There can be multiple factors that change with time and may even be dependent on each other. Such datasets are called Multivariate Datasets.

**Importance Of Time Series Analysis**

Well by now we have a basic understanding of what Time Series Dataset means and how it can be used. Below are the two major objectives of Time Series Analysis.

- Identifying Data Patterns
- Predicting from the Insights

Identifying Data-Patterns involve understanding how the variables are related and this understanding of relationship is used to predict future values.

Based on the example of Sales Forecasting given above, specific patterns are recognized from the actual recording and are formulated into relations that would in turn help in making predictions for upcoming months or the same months for the next year.

**Models that best fit the Time Series**

There are numerous Data Science and Machine Learning libraries that incorporate many models to solve Time Series Analysis problems in Python and R.

Some of the most common models used for Time Series Analysis are:

- ARIMA models
- Box-Jenkins Multivariate Models
- Holt-Winters Exponential Smoothing
- Unobserved Components Model

**Conclusion**

Time Series Analysis is one of the most common Data Analysis problems that exist. There are several models that fit to serve the Time Series Analysis problems efficiently and tools that offer these models. Time Series Forecasting is employed in a number of real-life applications such as:

- Economic Forecasting
- Marketing and Sales Forecasting
- Yield Projections
- Seismological Predictions
- Military Planning