InputParams class - FLAGR - Rank Aggregation Library

The InputParams class stores options and execution parameters that have been passed to FLAGR by the user. These parameters concern input and output file locations, rank aggregation methods, algorithm hyper-parameters, etc. See the table below for a complete list of the supported parameters and their respective valid values.

This object is passed as an argument to multiple functions of FLAGR including the implementations of the rank aggregation methods. This is how these implementations get access to the user-defined hyper-parameters.

Implementation files
The InputParams class is defined in the src/InputParams.h header file; its member functions are implemented in src/InputParams.cpp.

Technical Details
The supported parameters include:

Parameter	Data Type	Description
`input_file`	String (ASCII)	The path of the input CSV file that contains the preference lists to be aggregated.
`rels_file`	String (ASCII)	The path of the optional CSV file that contains the relevance judgments of the input list elements. If set, it automatically triggers the evaluation process of the generated aggregate list. Otherwise, no evaluation takes place.
`output_file`	String (ASCII)	The file system location where the output file with the aggregate list will be stored. If left empty, the default OS temp directory is used.
`eval_file`	String (ASCII)	The file system location where the output file with the evaluation of the aggregate list will be stored. If left empty, the default OS temp directory is used.
`random_string`	String (ASCII)	A string that is embedded in the names of the output files. Used when FLAGR is compiled as a shared library.
`aggregation_method`	Integer	Determines the algorithm that will be used to perform rank aggregation. The valid values are: `100` for CombSUM with Borda normalization `101` for CombSUM with Rank normalization `102` for CombSUM with Score normalization `103` for CombSUM with Z-Score normalization `104` for CombSUM with Simple Borda normalization `110` for CombMNZ with Borda normalization `111` for CombMNZ with Rank normalization `112` for CombMNZ with Score normalization `113` for CombMNZ with Z-Score normalization `114` for CombMNZ with Simple Borda normalization `200` for the Condorcet Winners method `201` for the Copeland Winners method `300` for the Outranking Approach `400` for Kemeny optimal aggregation `401` for Robust Rank Aggregation (RRA) `5100` for DIBRA using CombSUM and Borda Normalization `5101` for DIBRA using CombSUM and Rank Normalization `5102` for DIBRA using CombSUM and Score Normalization `5103` for DIBRA using CombSUM and Z-Score Normalization `5104` for DIBRA using CombSUM and Simple Borda Normalization `5110` for DIBRA using CombMNZ and Borda Normalization `5111` for DIBRA using CombMNZ and Rank Normalization `5112` for DIBRA using CombMNZ and Score Normalization `5113` for DIBRA using CombMNZ and Z-Score Normalization `5114` for DIBRA using CombMNZ and Simple Borda Normalization `5200` for DIBRA using the Condorcet Winners method `5201` for DIBRA using the Copeland Winners method `5300` for DIBRA using the Outranking Approach `600` for the Preference Relations Method `700` for the Agglomerative Aggregation Method `801` for Markov Chains 1 (MC1) `802` for Markov Chains 2 (MC2) `803` for Markov Chains 3 (MC3) `804` for Markov Chains 4 (MC4) `805` for MCT
`correlation_method`	Integer	The correlation method that is used to measure the distance between an input list and the temporary aggregate list. The valid values are: `1` for the Spearman's \(\rho\) correlation coefficient. `2` for the scaled variant of Spearman's Footrule distance. `3` for Cosine similarity of the lists' vector representations. `5` for the Kendall's \(\tau\) correlation coefficient.
`weights_normalization`	Integer	The voter weights normalization method. Used when the DIBRA algorithm is selected. The valid values are: `1` for no voter weight normalization. `2` for normalizing the voter weights with min-max scaling. `3` for z-normalizing the voter weights.
`max_iterations`	Integer	This parameter controls the maximum number of iterations. FLAGR will stop the execution of DIBRA if the requested number of iterations have been performed, even if the voter weights have not fully converged.
`max_list_items`	Integer	Limits the length of the input preference lists.
`eval_points`	Integer	Determines the elements in the aggregate list on which the evaluation measures (i.e. Precision, and nDCG) will be computed. For example, for `eval_pts=10` FLAGR will compute \(P@1, P@2, ..., P@10\) and \(N@1, N@2, ..., N@10\).
`list_pruning`	Boolean	Triggers a weight-dependant list pruning mechanism. Used in combination with the DIBRA weighted method only.
`convergence_precision`	Float or Double	Controls the convergence precision. This tolerance threshold represents the minimum precision of the difference between the voter weight in an iteration and the voter weight of the previous iteration. Used in combination with the DIBRA weighted method only.
`alpha`	Float or Double	The \(\alpha\) hyper-parameter of the Preference Relations method.
`beta`	Float or Double	The \(\beta\) hyper-parameter of the Preference Relations method.
`gamma`	Float or Double	The \(\gamma\) hyper-parameter of DIBRA.
`c1`	Float or Double	The \(c_1\) hyper-parameter of the Agglomerative Aggregation method.
`c2`	Float or Double	The \(c_2\) hyper-parameter of the Agglomerative Aggregation method.
`pref_thr`	Float or Double	The preference threshold of the Outranking Approach. It takes values in the range \([0,1]\).
`veto_thr`	Float or Double	The veto threshold of the Outranking Approach. It takes values in the range \([0,1]\).
`conc_thr`	Float or Double	The concordance threshold of the Outranking Approach. It takes values in the range \([0,1]\).
`disc_thr`	Float or Double	The discordance threshold of the Outranking Approach. It takes values in the range \([0,1]\).

The following interactive block diagram depicts the functionality of the InputParams class inside the entire architecture of FLAGR.

3.6. InputParams class