Acensify anomaly detection
- #Computer Vision
- #Smartassistant
A virtual assistant which analyses user input and automatically detects anomalies based on other records within ERP systems with the target of reducing input errors and minimising correction effort.
- Natural Language Processing
Impact
The solution allows to reduce the reduction of input errors and minimises correction efforts in ERP systems.
- Automated anomaly detection for improved accuracy
and efficiency in data analysis.
- Enhanced decision-making processes.
- Reduced operational costs.
Services we provided
PoC of Smart assistant for anomaly detection (values in columns that significantly deviate from previous observations) in data imputed, including
Detecting anomalies and assistant fine-tuning.
- Covering most of the functions of the smart assistant.
- Hardcoding specific simple rules for special columns.
- Fine-tuning anomaly coefficients and smoothing the work of the assistant.
Automating detecting anomalies that span across multiple tables
- Relationships analysis
- Algorithm development
- Algorithm implementation.
Tech Stack
Python
Distribution Analysis
Detecting Anomalies
ML Approaches
Challenges and Solutions
🧐 Challenges
- To develop an algorithm that can effectively analyze user input and detect anomalies based on other records within ERP systems.
- To fine-tune the assistant’s coefficients and rules to ensure accurate anomaly detection.
💡 Solutions
During our work, we accomplished the following:
- Conducted data analysis
- Developed a custom unsupervised learning algorithm
- Applied our algorithm to the tabular data for accurate anomaly detection.