Getting Started
Nexosis Concepts
Forecasting
Anomaly Detection
Security
Troubleshooting
Algorithms Employed
Our machine learning platform has a growing suite of about 30 core algorithms with a total of over 300 permutations. Given your data, our suite of algorithms will tune and select a model that best fits your data and the problem you’re trying to solve.
Forecasting a future quantity
Solves the question, how much of x can I expect in the future?
- Algorithms support hourly, daily, weekly, monthly, and annual seasonalities
- All support Anomaly Smoothing and Model Ensembling
- ARIMA
- Various combinations of
- AutoRegressive component with p parameters
- Differencing component with d parameters
- Moving Average component with q parameters
- with external regressors
- Exponential Smoothing
- Simple
- Double
- Triple
- with Box-Cox Transformation
- Autoregressive Neural Network
- with or without external regressors
- Multiple Linear Regression
- with or without external regressors
- Spline
- Seasonal and Trend Decomposition using Loess
- with ARIMA
- with or without external regressors
- with Exponential Smoothing
- Bayesian Time Series Regression
- with or without external regressors
- Additive Model
- Home-grown Nexosis Algorithms
Predicting a variable
Solves the question, what can I expect x to be?
- Least Squares
- Linear
- Polynomial
- Elastic Net
- Lasso
- Ridge
- Support Vector Regression
- Linear Kernel
- Polynomial Kernel
- Radial Basis Function kernel
- Sigmoid Kernel
- Multi-Layer Perceptron (Neural Network)
- with 1, 2, or 3 hidden layers
- Rectified Linear Unit Function
- Hyperbolic Tan Function
- Sigmoid Function
- Random Forest
- K-Nearest Neighbor
- Logistic Regression
- Naive Bayes
- XGBoost
