Introducing Obviously AI Time Series

Product
Data Literacy

With the launch of our sweet new Time Series product, time series analysis and forecasting just got a whole lot easier.

Ready to predict sales, demand, or visits down to a specific date—with the least data possible? Starting today, with the launch of our sweet new Time Series product, time series analysis and forecasting just got a whole lot easier.  

Here’s everything you need to know about Time Series: What it is, why it’s a difficult problem to solve, and how Obviously AI makes it faster and easier for you. 

What is Time Series 

First things first: What is time series?

Time series is a type of data that records events happening over a period of time. We analyze the series to determine a long term trend to forecast for the future. 

There are two ways to leverage time series:

  • ​​Time series analysis comprises methods for analyzing time series data in order to extract meaningful statistics and other characteristics of the data. This field of study seeks the “why” behind the dataset.
  • Time series forecasting is the use of a model to predict future values based on previously time-dependent observed values.

Time series data is different from cross-sectional data, which observes multiple factors that may or may not be time dependent at a single point in time. 

While time plays a role in standard AutoML datasets, a time series dataset is different because it adds an explicit order dependence between observations: a time dimension. A time series is a sequence of observations taken sequentially in time. 

A time series dataset, therefore, captures each and every change to the system/process/behaviour, and the output from it can be used to forecast all kinds of things.

We specifically focus on univariate time series forecasting, such as where we concentrate on a single numeric column that is time dependent.

For instance, time series forecasting can be used to predict when customers will make their next purchase. This helps businesses make decisions about when to launch new products or when to send their customers updates about their new product. 

This is why we’re so excited about our newest product: in just a few clicks, you can effortlessly predict business critical events down to a specific date with the least data possible, without writing code.



Why Time Series is a Difficult Problem to Solve

Time Series is one of the most difficult problems to solve in data science because you need to look back at historical data to try and guess accurately what could happen next. 

One of the challenges with Time Series is how to formulate a problem: do we want to forecast a precise value? Do we want to anticipate what will happen next? Instead of forecasting a value, do we want to extract a trend? 

All of these are very different problems. They all require different types of models to identify the right approach to solve these problems and the right algorithm to provide you with the highest possible accuracy. 

Doing all of that is very complex and time consuming for a data science team. But the good news? We just made it incredibly easy. 

Like what you see? Be sure to vote for us over on Product Hunt!

How Obviously AI Makes Time Series Faster and Easier 


Let’s explore what Time Series looks like in Obviously AI by walking through a use case. 

In this instance, we would like to forecast sales over a 14-month period. The dataset we’re using has monthly car sales data for the past 8 months. We’ll use this information to predict car sales. 

Just like AutoML, it’s important to have a dataset tailored accordingly so that we can generate accurate predictions within minutes. But for time series, we don’t need much preparation; everything is handled automatically within the platform. 

First, you’ll need to choose the data source. We’re uploading a CSV file for our datasource, but you can also integrate a database to the platform.


Choosing a data source for Time Series in Obviously AI

To make predictions, our dataset will need:

  • A minimum of 12 rows 
  • One date column
  • One prediction column
  • As less missing values as possible

All of our data is automatically pre-filled for the date column and prediction column. Seasonality will automatically be filled by platform depending on the data level chosen of our data. Once we define the values we want to analyze (Data level = month, aggregation function = sum), we simply click “Start Predicting” and we’re on our way!

To view a more in-depth walkthrough of  how to build a model for time series, be sure to watch our video on Predicting Sales video.


Defining the predictions for Time Series in Obviously AI

In seconds, Obviously AI trains and builds your model and in less than a minute, you’ll have a prediction report to explore. As you can see in the image below, there are five main tabs:

  • Overview: provides details on the performance of the model built
  • Predictions: allows you to run forecasts
  • Export: helps you export and share your trained time series model with colleagues
  • Advanced Graphs: visualizes the performance of the trained model on the test data
  • Tech Specs: provides you with an in-depth detail on advanced model metrics and lets you run other time series models incase you are interested in

A graph showing the distribution of Time Series dataset in Obviously AI

So, we’ve collected all our historical observations and we’ve successfully predicted what is likely to happen if these trends continue. What is this forecast telling us?

For starters, we can see when sales will be at their peak and we can see when sales will be low. This kind of data is valuable. Human memory is prone to error. Managers might think their sales peak at Christmas and bottom out in February—but is that really true? Having this kind of information removes doubts. 

We can also use this information to inform decision-making. For instance, managers can optimize their inventory stock. No undershooting or overshooting their estimates. With this real-time data in hand, managers can more easily prepare for times ahead.


Time Series Predictions Plot graph in Obviously AI

You’ll notice that all our graphs are interactive and provide details of each datapoint as you hover over them. You can also zoom in/out and expand a particular region on the graph, giving you more insights into the predictions generated by the model.

We can export this data via CSV by clicking the download button within the predictions table.

And under the Export tab, there is an option to share the link as a webapp with anyone on the team! 


Export Time Series as Web App in Obviously AI

This means anyone who gets this link can now use the model we built to start making their own predictions. If your team doesn’t know what Obviously AI is or how it works, anyone can now run their own time series forecasts.

So, if a colleague wants to run a 12 month forecast instead of the 14 months forecast we ran, they can do exactly that without having an Obviously AI Account.

You can also export your time series model as an API by copying the source code provided on the Export tab and run predictions.

Related: Our Guide to Machine Learning


What Time Series Can Help You Do

Times Series removes doubts and best guesses from decision-making. Knowing when to expect a lull in sales or a surge in demand provides incredible value to any company. 

We’ve designed Obviously AI to work with less data while giving the most accurate predictions in minutes, without writing code. This means anyone can leverage Time Series to easily and quickly make predictions and decisions.

Obviously AI Time Series


For instance, an accurate forecast of the stock index could allow investors to understand an overall trend of the market. With that information, they’re able to make informed trading decisions.

Time Series has so many use cases and is used all over the world. You can leverage Time Series to help predict:


Related Reading: Using No-code AI to Predict Hospital Readmission


With the power of Time Series, you have the ability to predict business critical events down to a specific date with the least data possible. When you have the ability to predict the future, you have the power to think proactively. Which means you’re ready for anything

And, when that power is coupled with the no-code machine learning, you’re not only able to forecast, you’re able to forecast in a matter of seconds. 

Putting it All Together 

We’re really excited about this new feature—Obviously AI clients can now quickly and easily understand seasonal trends, dig deeper into them to see why they occur, and predict the likelihood of future events. Time Series will help you think proactively, move with agility, and make decisions with confidence.  

So what are you waiting for? Go experience it yourself, and don’t forget to share your thoughts and feedback. Not ready to dive in yet? See it in action as we use Time Series to forecast stock prices.

And don't forget to vote for us over on Product Hunt!




Become the Data Scientist your team always needed.

Get Started

Get Started Now

See how no code machine learning can transform your business and change how you make decisions.