Readme - SunnerLi/Indoor-Locating GitHub Wiki

Indoor-Locating ( The Practice of UjiIndoorLoc Dataset )

Abstract

This repository try to build the model that can solve the indoor location problem. The dataset are downloaded from Kaggle. SVM, random forest, gradient boosting tree and DNN are adopted in this program. For the comparision, the random forest is the fastest and most powerful model for this problem.

Result

Model type	SVM	Random forest	Gradient boosting tree	DNN
Error value	58786.35	22390.67	45143.24	234203.36

Train by Yourself

You should download the data from Kaggle. Next, depress the file and put them in the same folder.

sunner@sunner-pc:~/loc/$ ls
abstract_model.py  data_helper.py  dl_model.py  main.py  ml_model.py  Readme.md  TrainingData.csv  ValidationData.csv

Notice

The code might not be accepted by Kaggle submission mechanism since it use some deep learning model. (For example, tensorlayer) What's more, the version of python is 2.7 while the 3+ version are accepted on the platform.