Lunchcat is a demo chatbot our team created that uses natural language processing (NLP) and helps you and your friends easily and quickly split lunch costs.
Digamma developed a machine learning based proof-of-concept project for the Silicon Valley-based company Happy Gears.
Driven by the notion that the online traffic and patterns of the future can be predicted by analyzing the traffic and patterns of the past, Digamma’s project for Happy Gears focused on the analysis of digital network traffic patterns and prediction of online network traffic flows.
The technology was intended for capacity planning groups at large companies that run their own networks and data centers such as Google, Facebook, Dropbox and related companies.
The machine learning components of the project included: Machine learning components of the project:
- Prediction of network traffic based on past data (e.g. time series)
- Prediction of changes of traffic in the whole network in response to local changes. Algorithm assumes multivariate Gaussian distribution of data, and fits the model using Singular Value Decomposition (SVD)
- Anomaly detection and investigation: finding correlations of anomalies with other variables
Machine learning techniques used:
- Calibrated seasonal autoregressive integrated moving average (SARIMA) modelling
- Dynamic Time Warping (DTW)