if you want the project pls call @8125424511
Exploring Trajectory Prediction through Machine Learning Methods
In this paper author is describing concept to predict next location of single or multiple users by training trajectories (users previous location movement latitude and longitude) of their previous locations using RNN (Recurrent Neural Networks) advance version called LSTM (Long Term Short Term Memory) and Seq2Seq (sequence to sequences) algorithms. Predicting location of users plays an important role for 5G Internet networks as network service providers need to allocate nearest resources (cloud servers who take users mobile heavy computation task and process that request and send result back to mobile, if nearest cloud allocate to user then response will be faster and this nearest allocation can be done if users next locations can be predicted) to users to process their mobile request data.
Earlier algorithms such as KNN, SVM etc can predict user’s location but their performance will not be efficient when data size goes beyond limit. To overcome from this problem author is using combination of advance LSTM and Seq2Seq algorithms which is very much efficient in prediction and fast processing.
LSTM algorithm contains multiple copies of memory from training data and each copy consists of input, output and forget cells. Input contains input data and output contains output data and if output is not related or new output is better than old output then forgot cell contains old output data. This process continues till all training data allocated to input and output cells. New test location data will be applied on LSTM train output cell to predict future location.
Seq2Seq algorithms can be included inside LSTM algorithm which can help in predicting sequences of future locations from train data. Seq2Seq algorithm consists of two parts called Encoder and Decoder. Encoder will convert training data into two dimensional array and Decoder will predict next sequences from those two dimensional array.
To implement above concept author is using Geolife real life trajectory movement dataset which consist of user’s movement latitude, longitude and users id and each user has 9 locations. By training this dataset with LSTM and Seq2Seq we can predict next sequences of user’s locations.
thank you for your comment
pls call me on 8125424511