Software | Tangent Works - Advanced Forecasting

OVERVIEW

Meet TIM

TIM , Tangent Information Modeler, is a predictive modeling engine that automates the forecasting and anomaly detection process by analyzing time-series data and generating accurate models based on the patterns it detects.

TIM can be deployed as a standalone Cloud-based Web application or as a
Web service that can be easily integrated with your existing business
applications through a simple API. Your TIM-enabled end to end solution
can be deployed entirely in the cloud, on premise, or using a hybrid
approach.

To build a model, all you need to do is feed your historical data into TIM, and
in just minutes it creates a predictive model ready for validation and
deployment.

KEY FEATURES

Other modeling tools require tedious, manual “feature engineering” that requires a domain expert to select which combinations of input data over what time delays are potential predictors and are relevant to the machine learning process. This compute-intensive trial-and-error process can take weeks, since the number of combinations grows exponentially with the number of input variables and time intervals.

TIM is different. It automates the feature engineering process by analyzing the historical input data and determining which data is relevant to the predicted output. For example, if a machine’s vibration is sensitive to condensation, which is related to a hot, humid day following a cool day- TIM will spot this pattern and incorporate it into the model.
Seeing is believing. Many of our customers had existing models that were built and tuned manually with months of effort. Imagine their surprise when TIM created a model in minutes with equivalent or even better accuracy. It’s no wonder that Tangent Works was a winner at the recent GEFCom 2017 competition for energy demand forecasting accuracy.
TIM’s underlying technology is extremely compute-efficient and therefore can process huge data sets in minutes. That means you can automatically create models for hundreds or even thousands of individual assets or customers and update those models as frequently as you’d like. TIM also helps your organization scale to handle the increasing demand for predictive models by automating the most tedious aspects of model building and fully leveraging your existing resources.
A challenge with any predictive model is how to explain it to the people who depend on it for forecasting.

TIM creates its models in human-readable format, as both an equation and as a visual map that shows the weighting and relationships of all the input variables. For example, you’ll see that the machine vibration is directly related to the sequence of hot, humid days and cool days, but is also related to the time since the last routine maintenance.
TIM was architected for rapid integration with your existing databases, BI tools and other enterprise applications since all of its functions are easily accessible through its REST API. It also gives you the utmost deployment flexibility with support for cloud-cloud, on-premise and hybrid configurations. Most integration projects take just a few days to complete.

InstantML

InstantML for timeseries data offers Feature engineering, Model building and Model deployment in one go.

TIM creates models in seconds based on target and predictor series time series data with just one pass through the data. These models are made available through a single API, thus saving organisations the hassle of creating separate APIs for each of the models.

Traditional Machine Learning relies on hand crafted models.

AUTOML has introduced to bring automation in the development process and model selection, still this is an engineering technique.

TangentWorks introduces InstantML to automate Feature Engineering, Model Building and model deployment. With RTInstantML the model is based on available data at the specific moment and forecasting or anomaly detection is done in one step. This is completely new paradigm and allows forecasting and anomaly detection in seconds.

Gartner claims: "By 2025, 50% of data scientist activities will be automated by artificial intelligence, easing the acute talent shortage” We are convinced that our InstantML his technology allow companies data scientists to focus on where they can make real value instead of handcrafted models.

TIM’s breakthrough technology

Any time a software product automates a tricky manual
process, the questions comes up-how does it do that?

There’s a lot of buzz about AI these days, and technologies like Deep Neural Networks (DNN) or Convolutional Neural Networks (CNN) are often mentioned in the news. While these techniques are great for image classification and speech recognition applications, they don’t work very well on time-series data that is the basis for energy forecasting, anomaly detection and predictive maintenance. Applications where TIM really shines.

TIM’s “secret sauce” is based on a field of mathematics known as “information geometry”, which was originally developed in Russia and Japan in the 1950s-1980s, but only recently put into practice by Tangent Works. You’d have to be a research mathematician to understand the underlying concepts, but if you’re curious about terms like “Riemannian manifold”, “Bayesian inference” and “Tangent space”, you can look it up on Wikipedia. The real proof of TIM’s technical approach is that it just works. Hence the company name, Tangent Works.

Advanced forecasting and anomaly detection

Automatic model generation for time-series goes hand in hand with anomaly detection. TIM generates high quality models over the in-sample training data. Moreover, TIM’s anomaly detection module estimates distribution density functions of each explanatory variable and checks whether the incoming values in production (out-of-sample) have the same characteristics. If not a data / anomaly warning is dispatched to a user. The anomaly detection module can also be used in a standalone scenario to detect anomalies in your data.

Take a Tour

See for yourself what TIM is capable of and what it can mean
for your business.

Slide 1

Prepare your data file with the target variable and predictor candidates

Slide 1-1

Use the Web user interface to select a file to upload

Slide 1-2

TIM™ shows you the available time range for the target variable- in this case, 1 year

Slide 1-3

TIM also shows you all the potential predictors in your data set, e.g. humidity, temperature…

Slide 2

Click “Build” to generate a model for the selected target variable.
TIM then analyzes your data and creates the optimal model in seconds using a mathematical approach called information geometry

Slide 2-1

TIM calculates the mean absolute percentage error (MAPE) for your model, which is a measure of its prediction accuracy

Slide 3

View a visual map for the model showing the relative weighting and relationship of the predictors with the target variable.
In this example, need help to explain what two of the boxes mean

Slide 3-1

View the summary of the forecast in Microsoft Power BI or any other data analysis tool

Slide 3-2

Explore the model and data

Slide 3-3

View the predicted target value and predictors over the specified future time period

Slide 4

Your model is now ready for deployment as a standalone Web interface or integrated with your existing IT applications

Step 1

Load the data into TIM.

Step 2

Automatically build an accurate model.

Step 3

Understand the model with TIM’s “explainable AI”.

Step 4

Deploy the model in your predictive analytics application.

Deployment

TIM can be deployed in various ways to meet your
requirements. You can deploy TIM in the cloud or on-premises
or even in the edge on IoT devices.

Microsoft Azure Cloud

TIM can be used as a cloud service managed by Tangent Works. This offers you ease of use and deployment.

Your Azure Cloud

TIM can be deployed on your own Microsoft Azure subscription.

Your Infrastructure

TIM engines and applications can be deployed on your own IT infrastructure. The state of the art container-based architecture allows for easy installation and integration in your environment with minimal effort.

IoT Edge

TIM can be deployed on various IoT edge devices. This allows you to bring model building, advanced forecasting and anomaly detection to the field.

Request a Demo

Interested in trying TIM out?

Get in Touch

Let’s explore together how you can get more out of your data.