Modeling and Monitoring of Indoor Populations using Sparse Positioning Data

Implementation Details

The implementation concerns 4 folders which are:

• preprocess: generate probabilistic populations.
• estimators: population prediction models
• input: population dataset
• query: CMPP query

Requirements

• Pytorch 1.8.1
• Numpy 1.19.2
• Pandas 1.1.3
• Sklearn 0.24.1
• Matplotlib 3.3.2
• networkx 2.8.4

You may use " pip3 install -r requirements.txt" to install the above libraries.

Usage

To generate probabilistic populations:

cd ./preprocess; nohup python3 -u generate_populations.py > population.log &

To train and test population prediction models (e.g., ME):

cd ../estimators/; nohup python3  -u ME.py --epochs 500 --batch_size 64 --time_interval five_mins > ME.log  &

To run CMPP query (e.g., varying \theta):

cd ../query; java -cp CMPP-0.0.1-SNAPSHOT.jar experiments.Experiment_Pop

or you may run it in IDE (e.g., IntelliJ IDEA). Multiple csv files will be generated to record the f1-score, response time, and memory usage respectively.

Explaination of Parameters

time_interval: prediction time interval (i.e., \delta)

batch_size: the number of samples for back propagation in one pass

epochs: the number of training rounds

Datasets

The whole datasets are very large and thus moved to Google Drive.