Create README.md

code/README.md

@@ -0,0 +1,52 @@
+# HowTo
+
+This folder contains additional code used for model testing and feature generation. We assume the experimenter has run an experiment using the WiSHFUL UPIs for WSNs that support TAISC (http://www.wishful-project.eu/software) and a global control program with event-based monitoring using the *RIME_appPerPacket_rxstats* event.
+
+The generic flow for modeling the MAC-level performance predictor is described below.
+
+## Data collection
+
+To generate experimental data for MAC statistics the following steps are involved:
+* Create WiSHFUL control program (configure network and radio parameters)
+* Register events for which data has to be collected (MAC statistics)
+* Define scenario
+* Define monitoring duration
+* Dump data for post-processing (SQLite, CSV, MySQL, OML...)
+
+![datacollection](https://cloud.githubusercontent.com/assets/7999611/22187153/9311db1e-e101-11e6-889e-95d2bbc69e05.JPG)
+
+
+## Feature Generator
+
+### *Offline Feature Extraction*
+
+To extract MAC-level performance statistics from the collected the *MACperfFeatureGenerator.py* script is used. The script takes the path to the experimental data as input argument.
+```
+python MACstatsFeatureGenerator.py -d “Path to data”
+```
+![pre_process](https://cloud.githubusercontent.com/assets/7999611/22187164/c1465528-e101-11e6-9f2d-ab94836bf3f0.jpg)
+
+### *On-the-fly Feature Extraction*
+
+MAC-level performance statistics can be automatically extracted on-the-fly in the data collection phase. The framework extension module will be soon available. 
+
+![otf_featuregen](https://cloud.githubusercontent.com/assets/7999611/22334205/223a24ba-e3da-11e6-8f50-8c96124d5129.jpg)
+
+
+## Create MAC-level performance predictor
+Train the MAC-level performance predictor:
+```
+java -classpath weka.jar weka.classifiers.functions.MultilayerPerceptron -t “Path to dataset" -L 0.1 -N 2000 -H 10 -d “Path to model"
+```
+![genmodel](https://cloud.githubusercontent.com/assets/7999611/22187171/cf2cda7c-e101-11e6-8684-41ebd0e7b8d4.jpg)
+
+
+## Evaluate Model
+
+To evaluate the provided serialized neural network model with trace-based simulation, the testModel.py script can be used in the following way:
+```
+python -p "Path to Weka application folder" -m "Path to Weka serialized model object" -d "Path to testing set"
+```
+The script generates a performance graph showing the real vs. predicted instance.
+
+![testmodel](https://cloud.githubusercontent.com/assets/7999611/22187173/d9e6da58-e101-11e6-973f-8b4ad5bd3a03.jpg)