- Update to use
dplyr::n()instead of justn().
- Changed solution vector to be integer, not numeric, to reduce space - #42
- Removed
return_earlyargument frommax_coverage. - Refactored the solving process for
max_coverage
max_coveragenow has consistent output for solverglpksummarise_coveragenow usesdistance_cutoffinstead ofdist_indiccoveragenow propagates distance cutoff properly now has consistent output for solverglpk
- Now byte compiles by default (which may not be necessary for R >= 3.5)
max_coveragenow usesglpkby default, as it is much faster.
- This release is created to assist in replicating previous journal articles that used this software. The API of maxcovr is likely to change from here.
max_coverage_relocationnow works for different data, thanks to a bug fix.- prototype solver engines added. maxcovr now works with
glpk, andgurobi, in addition tolpSolve. Testing still to be conducted. - summary function
coverageadded as a one liner to take two dataframes and from the coverage of one dataframe on another. Documentation and tests still need work summary_mc_cv_relocateadded forrelocationmethods. This will eventially be an S3 method, I think?- rewrote the extraction functions for results from
max_coverage, replacing them with key functions. This will make things easier to debug and extend in the future. These functions start withextract_. There will likely be some updates to this in the future - started work on refactoring the
max_coveragefunctions to work with vectors, this fule is calledmaxcovr-refactor-vectors. This provides some more efficient computation of preparation of matrices for optimisation, and will provide substantial speedups for larger N, and for when multiplen_added's are needed. max_coveragenow returns the entire solution fromlpSolve, except for the constraints, because they are too large (can easily be over 1Gb)
summary.maxcovrandsummary.maxcovr_relocationmethod now return information about the previous distances.- Improvements have been made to the overall returning API of
max_coverage_relocationso that it plays better with extracting summary information.
- Removed the scraggly test dataset that was lurking in the shadows and making my model not work.
- new function,
max_coverage_relocation, takes a arguments for total cost, installation cost, and relocation costs and then works out how many facilities it can place, and potentially remove and replace to obtain optimum coverage. The function is currently Under development. In the future it will be absorted intomax_coverage. - added a results extraction method for
max_coverage_relocation,extract_mc_result_relocation. Eventually this will be be absorbed into theextract_mc_resultfunction, through some kind of S3 method. - added
printandsummaryS3 method formax_coverageandmax_coverage_relocation, and anis.maxcovr_relocationandis.maxcovr, which should be handy for testing.
- Changed the specification of m_under_i and m_over_i to more clearly reflect "The gain of removing an AED from position i"
- Removed the scraggly test dataset that was lurking in the shadows and making my model not work.
- added a vignette that describes how to perform cross validation on the maxcovr problem using modelr and purrr
- maxcovr now returns a tibble of lists. This makes it easier to perform other tasks, and more consistent and easier to work with.
- added c++ functions to calculate the distance matrices and binary matrices
- added wrapper function
nearestto find the nearest lat/long points from one dataframe to another and then calculate the distance between the two. This is at least 10 times faster than the previous method using joins and dplyr. - changed API for max_coverage, it now takes 3 dataframes, but it does more of the behind the scenes work, at the cost of taking slightly longer as it calculates the A matrix. Speedups are definitely still available though, as the c++ functions are yet to be parallelised.
- max_coverage now also calculates a summary, which can be retrieved as
mc_model$summary.
- fixed c++ code for distance_matrix
- added
summarise_coverage()function, which calculates coverage based upon a distance dataframe created fromfacility_user_dist(). - added
york_crimedata containing crime locations around the city of york.
- improved README to include examples of usage (again)
- removed artifacts relating to previous work; maxcovr now works for other kinds of data
- renamed from
coperturatomaxcovr - added
yorklisted building data, taken from data.gov.uk
- gurobi solver in max_coverage.
- faster matrix methods in max_coverage
- added draft cpp code for
spherical_distanceanddistance_matrix - updated
max_coverageto include theextract_mc_resultinside the function, rather than needing a separate function call. This was to save space, as the A matrix and friends can be rather large, especially if you are running the coverage multiple times under different conditions.
- fixed bug in spherical_distance to return metres. Before this is was assuming there were 100m in a kilometer!
- added parameters to
facility_user_distance-coverage_distance, which allows you to specify the coverage you are interested in, andnearest, where you specifynearestto be "facility" if you want the nearest facility to each user, "user" if you want the nearest user to each facility, and NULL if you just want the complete pairwise distances.
- Added a
NEWS.mdfile to track changes to the package. - Added a helper function to assist with extracting the results from the model
- This will eventually be broken up into smaller functions, so that you can be really specific about what function you want to run to get what result. Ideally, each function, except for the one that runs the model, should return one, or maybe two arguments.