yangxhcaf/drsstl

Divide and Recombined for Spatial-Temporal Analysis

drsstl: Divide and Recombined for Spatial Seasonal Trend Loess

This is an R package that provide a way to analyze spatial-temporal data based on SSTL (Spatial Seasonal Trend decomposition using Loess) routine under the divide and recombined framework.

Installation

# from github
devtools::install_github("XiaosuTong/drsstl")

Fitting

The main function to call in the package is drsstl function, which is a wrapper function of sstl_mr and sstl_local two functions. drsstl can be used to carry out SSTL routine on the dataset either in the memory or on the HDFS.

Prediction

The function for prediction is predNewLocs function, which is a wrapper function of predNew_mr and predNew_local two functions. Depends on where the original fitting results are saved, the prediction at new locations will be conducted either in local memory or on HDFS using MapReduce.

Data

In the data subdirectory, tmax.RData inlcudes a data.frame named tmax_all which can be used as example to illustrate the usage of drsstl in local memory. station_info.RData contains a data.frame which includes all stations information. It is usefual for both local and mapreduce situation of drsstl. User has to copy the station_info.RData to HDFS if the mapreduce is chosen.

In inst subdirectory, tmax.txt is the raw data file contains the maximum temperature data. User can copy it to HDFS and conduct drsstl under divide and recombined framework.

Acknowledgment

drsstl development is depended on

• stlplus package by Ryan Hafen
• datadr package by Ryan Hafen
• Rhipe package by Saptarshi Guha
• Spaloess package by Xiaosu Tong
• loess function in base R by R core Team