-
Notifications
You must be signed in to change notification settings - Fork 90
Home
Avenir provides a set of machine learning solutions on Big Data platform targeted towards solving certain business problem. Unlike other machine learning packages, Avenir has a very top down business use case driven approach.
It provides solutions in the following areas
- Exploratory analysis
- Classification and prediction
- Sequence mining and prediction
- Reinforcement learning
As a business owner you need to ask yourself the following questions before choosing a data mining package
- Will I be able to quickly and easily find the solutions for my business problems.
- Will I be able to deploy the solutions without excessive CAPEX
- Will I be able to run the solutions in real time ,if necessary
- What kind of input data is needed
- How configurable is the system
- How does the system scale with increasing load
- How seamlessly does it integrate with exiting IT infrastructure
- Is it cloud ready
- Is there a convenient licensing model
- Once deployed, how do I know that the system is working and I have ROI
Avenir is an open source machine learning running on a Big Data platform comprising of Apache Hadoop, Apache Storm and Redis. Avenir has satisfactory answer to all the issues raised above as outlined below.
Avenir has a top down. Approach. For a set of business problems, it’s easy to find appropriate solutions and algorithms.
Avenir can be deployed In Amazon or any other public cloud without major upfront expenses.
Some of the predictive solutions are more appropriate in real time. Real time predictive solutions run on Apache Storm. Typically Hadoop is used to build a predictive model and actual prediction algorithms run on Storm.
Input data is in easy to understand CSV format. Type of data needed depends on the problem and the algoritms being used.
Avenir believes in giving power in the hands of the end users. There are many configuration parameters to control the behavior of the solution.
Avenir runs on an open source Big Data platform consisting of Hadoop, Storm and Redis. This platform is horizontally scalable with commodity machines.
Avenir stays away from complex file format and uses CSV file format for input and output. As long as the external enterprise system can generate CSV output to be consumed by Avenir and can consume CSV data generated by Avenir, there is no additional work necessary for integration.
Avenir can easily be deployed on Amazon or Google cloud. This approach makes more sense if the existing enterprise solution is already hosted on cloud.
Avenir has dual model. The community edition is free with software, scripts, tutorial and plenty of blogs for background technical material. The enterprise edition includes web based admit tool, support, training , consultancy and documentation.
Avenir provides Hadoop based tools for correlations and feature subset selection for classification problems. An example will be determination of critical parameters in predicting hospital readmission of patients.
When certain assumptions hold, discriminant based algorithms provide simple classification solution.
Naïve Bayes classifier is a proven and time tested classification tool. An example use case is prediction of customer churn.
It’s another proven and time tested classification tool. An example will be to predict whether a customer will be a repeat customer for a business
There are decision making tools. Given a set of choices, the system will try the different choices, learn from the rewards received from the environment and determine the most optimal choice. Some examples are deciding what product to sell, what advertisement to run etc.
These are predictive algorithms based on mining sequence using Markov chain model. An example will be determining the trajectory of customer loyalty. Another example will be personalized email marketing where the time to send the email is personalized based on behavior data.
There are predictive algorithms that work directly work on the data without building a model.