The objective of this project is to build a decision making process for a device registered to the Industry Marketplace to automatically choose the best service requests based on matching eCl@ass capabilities.
Please read instructions at https://www.hackster.io/naveenbskumar/autonomous-marketplace-device-adebf3 to setup environment.
The idea is to use the sent or rejected proposal data which were decided by human agent in the recent past. For this exercise we selected Decision Trees which are a non-parametric supervised learning method used for classification and regression. The goal is to create a model that predicts the value of a target variable by learning simple decision rules inferred from the data features.
We have considered Drone Connectivity Provioning service as an example.
To generate synthetic dataset use the script generate_data/index.js as follows.
$ cd generate_data
$ npm install
$ npx babel-node index.js > data.json
To build a Decision Trees model please use the notebook notebooks/Service_Provider_Decision_Tree_Classifier.ipynb. The model is serialized and saved to file using pickle module. The saved model is not portable across different python build so please build the model on the same machine or the machine with similar python build (32bit or 64bit). We used Raspberry Pi 4 for all setup.
The full example code which uses the model for decision can be found at industry-marketplace-python-helper/service_provider.py.
After starting up Service Provider and Service Requester servers, run following command to listen on CFP callback.
$ python3 industry-marketplace-python-helper/service_provider.py
The CFP request can be sent using UI by selecting "Drone Connectivity Provision" operation or running following command.
$ python3 industry-marketplace-python-helper/service_requester.py drone_connectivity_provision
Please set name and ip address/port in the service_requester.py and service_provider.py according to your set up.