Take your data
to the next level with
AI powered analytics
Adaptix.AI is an AI-Powered Analysis Toolkit for Time Series and Sensor Data, Accessible through REST API.
The AI tools have been carefully designed so that the system necessities minimal to zero configuration time by the user. Furthermore, you don’t need to be a data scientist or have any mathematical and machine learning skills to use it: Adaptix.AI tools are made so that they are self-trained. On the other hand, tweaking parameters is possible to do and the way they work is transparent to the user.
Whether you are an engineer, a web/app developer, a DevOps, or a data scientist, you may find in Adaptix an easy-to-use, zero-time-to-install, all-in-all AI solution that can resolve your major problems with time series and sensor data analysis and will synergize with your application or data platform to unravel the full potential of your data.
- Recognition of key patterns: the user queries and finds quasi-similar occurrences of important predefined patterns that are part of a sector-specific library (for example, recognising signal signatures of devices or specific events) delivered with the product.
- Custom pattern recognition : the user defines a pattern of reference (by providing an example or designing a curve) and the system queries for quasi-similar occurrences of this pattern in the past
- Automatic pattern identification and Trend extraction: The system queries the signal and automatically identifies trends and exceptional events occurred throughout a certain period.
- Real-time, short-term prediction: System learns from the past and predicts, at time t, the most favourable future scenario with a degree of confidence.
- Multi-Scenario prediction: Provide a number of probable scenarios with a respected degree of confidence.
- White box analytics: Provide justification for each of the aforementioned scenarios by highlighting similar simulations in the history of the stream.
Clustering & Segmentation
- Classification of data streams and according to different time bases: daily, weekly, monthly etc.
- Customized clustering between different types of entities (streams, devices, groups – e.g. floor or buildings) based on different parameters: stream key patterns, mean value per month, day…
The system listens to the stream and after some time it automatically detects anomalies and other complex patterns in the signal whenever they may occur. Three different modes of detection are supported:
- Supervised: user defines complex and abnormal patterns to be recognized automatically and in real-time.
- Automatic: System is trained as-it-goes without any supervision, progressively improving it’s capability to detect abnormal events.
- Semi-supervised: System learns actively; By incorporating user feedback in a simple yes-no form for every anomaly detection, the system accelerates it’s training process.
Adaptix also offers a pre-processing and transformation API for data streams. This feature is intended for cases where a prior transformation of the data is desired. Such transformations can include:
- Cleaning of raw data.
- Perform windowing and time-based aggregations.
- Combining multiple streams into one (e.g. computing the sum of different input data streams).