MedBootstrap Class Reference

Bootstrap wrapper for Medila Infrastructure objects, simplify the parameters and the input, output process. More...

#include <MedBootstrap.h>

Inheritance diagram for MedBootstrap:

Public Member Functions
void	parse_cohort_line (const string &line)
	parsing specific line.
void	get_cohort_from_arg (const string &single_cohort)
	A function which reads a single cohort definition from the command line and parses it.
void	parse_cohort_file (const string &cohorts_path)
	a function which reads cohorts file and stores it in filter_cohort.
	MedBootstrap ()
	defualt Ctor.
int	init (map< string, string > &map)
	Initialization string with format "parameter_name=value;..." each paramter_name is same as the class name field.
void	clean_feature_name_prefix (map< string, vector< float > > &features)
	cleans the initiale "FTR_" from the feature names in MedFeatures created by the infra pipeline
void	prepare_bootstrap (const MedFeatures &features, vector< float > &preds, vector< float > &y, vector< int > &pids, map< string, vector< float > > &final_additional_info, vector< int > &preds_order, unordered_map< int, vector< int > > *splits_inds=NULL)
	prepares the required vectors for bootstrap from MedFeatures &features
void	prepare_bootstrap (MedSamples &samples, map< string, vector< float > > &additional_info, vector< float > &preds, vector< float > &y, vector< int > &pids, vector< int > &preds_order, unordered_map< int, vector< int > > *splits_inds=NULL)
	prepares the required vectors for bootstrap from samples, additional_info
map< string, map< string, float > >	bootstrap (const MedFeatures &features, map< int, map< string, map< string, float > > > *results_per_split=NULL, with_registry_args *registry_args=NULL)
	Will run the bootstraping process on all cohorts and measurements.
map< string, map< string, float > >	bootstrap (MedSamples &samples, map< string, vector< float > > &additional_info, map< int, map< string, map< string, float > > > *results_per_split=NULL, with_registry_args *registry_args=NULL)
	Will run the bootstraping process on all cohorts and measurements.
map< string, map< string, float > >	bootstrap (MedSamples &samples, const string &rep_path, map< int, map< string, map< string, float > > > *results_per_split=NULL, with_registry_args *registry_args=NULL)
	Will run the bootstraping process on all cohorts and measurements.
map< string, map< string, float > >	bootstrap (MedSamples &samples, MedPidRepository &rep, map< int, map< string, map< string, float > > > *results_per_split=NULL, with_registry_args *registry_args=NULL)
	Will run the bootstraping process on all cohorts and measurements.
void	apply_censor (const unordered_map< int, int > &pid_censor_dates, MedSamples &samples)
	censors samples from samples based on time_range provided in pid_censor_dates.
void	apply_censor (const unordered_map< int, int > &pid_censor_dates, MedFeatures &features)
	censors samples from features based on time_range provided in pid_censor_dates.
void	apply_censor (const vector< int > &pids, const vector< int > &censor_dates, MedFeatures &features)
	censors samples from features based on time_range provided in pids,censor_dates.
void	apply_censor (const vector< int > &pids, const vector< int > &censor_dates, MedSamples &samples)
	censors samples from samples based on time_range provided in pids,censor_dates.
void	change_sample_autosim (MedSamples &samples, int min_time, int max_time, MedSamples &new_samples)
	changing the samples to be auto-simulations - taking max score in the time window for each pid
void	change_sample_autosim (MedFeatures &features, int min_time, int max_time, MedFeatures &new_features)
	changing the samples to be auto-simulations - taking max score in the time window for each pid
MeasurmentFunctionType	measurement_function_name_to_type (const string &measurement_function_name)
	convert measurement function name to type
Public Member Functions inherited from SerializableObject
virtual int	version () const
	Relevant for serializations.
virtual string	my_class_name () const
	For better handling of serializations it is highly recommended that each SerializableObject inheriting class will implement the next method.
virtual void	serialized_fields_name (vector< string > &field_names) const
	The names of the serialized fields.
virtual void *	new_polymorphic (string derived_name)
	for polymorphic classes that want to be able to serialize/deserialize a pointer * to the derived class given its type one needs to implement this function to return a new to the derived class given its type (as in my_type)
virtual void	pre_serialization ()
virtual void	post_deserialization ()
virtual size_t	get_size ()
	Gets bytes sizes for serializations.
virtual size_t	serialize (unsigned char *blob)
	Serialiazing object to blob memory. return number ob bytes wrote to memory.
virtual size_t	deserialize (unsigned char *blob)
	Deserialiazing blob to object. returns number of bytes read.
size_t	serialize_vec (vector< unsigned char > &blob)
size_t	deserialize_vec (vector< unsigned char > &blob)
virtual size_t	serialize (vector< unsigned char > &blob)
virtual size_t	deserialize (vector< unsigned char > &blob)
virtual int	read_from_file (const string &fname)
	read and deserialize model
virtual int	write_to_file (const string &fname)
	serialize model and write to file
virtual int	read_from_file_unsafe (const string &fname)
	read and deserialize model without checking version number - unsafe read
int	init_from_string (string init_string)
	Init from string.
int	init_params_from_file (string init_file)
int	init_param_from_file (string file_str, string &param)
int	update_from_string (const string &init_string)
virtual int	update (map< string, string > &map)
	Virtual to update object from parsed fields.
virtual string	object_json () const

Static Public Member Functions
static void	filter_bootstrap_cohort (MedFeatures &features, const string &bt_cohort)
	commit bootstrap cohort filter on a given matrix
static void	filter_bootstrap_cohort (MedPidRepository &bt_repository, MedModel &bt_filters, MedSamples &curr_samples, const string &bt_cohort)
	commit bootstrap cohort filter on a given samples
static void	filter_bootstrap_cohort (const string &rep, const string &bt_json, MedSamples &curr_samples, const string &bt_cohort)
	commit bootstrap cohort filter on a given samples

Data Fields
ROC_Params	roc_Params
	Controling the roc parameters: sensitivity, specificity...
Regression_Params	regression_params
	params for regerssion
Multiclass_Params	multiclass_params
	Controling the multi class parameters: top n...
map< string, vector< Filter_Param > >	filter_cohort
	the cohorts definitions. name to parameters range to intersect
map< string, FilterCohortFunc >	additional_cohorts
	not Serializable! additional cohorts given by function
float	sample_ratio
	the sample ratio of the patients out of all patients in each bootstrap
int	sample_per_pid
	how many samples to take for each patients. 0 - means no sampling take all sample for patient
bool	sample_patient_label
	if true will treat patient+label as the "id" for the sampling
int	sample_seed
	if 0 will use random_device
int	loopCnt
	the bootstrap count
bool	is_binary_outcome
	only used for validating bootstrap input
bool	use_time_control_as_case
	if True will use time window condition for controls same as cases.
bool	simTimeWindow
	Time window simulation (in cohorts with Time-Window filtering) - instead of censoring cases out of time range , treat them as controls.
float	censor_time_factor
bool	sort_preds_in_multicategory
size_t	num_categories
	number of categories
vector< pair< MeasurementFunctions, Measurement_Params * > >	measurements_with_params
	not Serializable! the measurements with the params

Static Public Attributes
static unordered_map< string, MeasurmentFunctionType >	measurement_function_name_map

Detailed Description

Bootstrap wrapper for Medila Infrastructure objects, simplify the parameters and the input, output process.

for more control and lower level interface please refer to bootstrap.h

Constructor & Destructor Documentation

MedBootstrap()

MedBootstrap::MedBootstrap

(

)

defualt Ctor.

look for ROC_Params defaults. cohorts consists of 1 cohort called "All" with not filtering

Member Function Documentation

apply_censor() [1/4]

void MedBootstrap::apply_censor	(	const unordered_map< int, int > &	pid_censor_dates,
		MedFeatures &	features
	)

censors samples from features based on time_range provided in pid_censor_dates.

the format is map from pid to max_date the after that date the sample is filtered.

Returns

update features - changes outcomeDate for controls to censor date.

apply_censor() [2/4]

void MedBootstrap::apply_censor	(	const unordered_map< int, int > &	pid_censor_dates,
		MedSamples &	samples
	)

censors samples from samples based on time_range provided in pid_censor_dates.

the format is map from pid to max_date the after that date the sample is filtered.

Returns

update samples - changes outcomeDate for controls to censor date.

apply_censor() [3/4]

void MedBootstrap::apply_censor	(	const vector< int > &	pids,
		const vector< int > &	censor_dates,
		MedFeatures &	features
	)

censors samples from features based on time_range provided in pids,censor_dates.

pids and censor_dates are same sizes. for each pid and the coresponding date in censor_dates, filtering pid's samples after that date.

Returns

update features - changes outcomeDate for controls to censor date.

apply_censor() [4/4]

void MedBootstrap::apply_censor	(	const vector< int > &	pids,
		const vector< int > &	censor_dates,
		MedSamples &	samples
	)

censors samples from samples based on time_range provided in pids,censor_dates.

pids and censor_dates are same sizes. for each pid and the coresponding date in censor_dates, filtering pid's samples after that date.

Returns

update samples - changes outcomeDate for controls to censor date.

bootstrap() [1/4]

map< string, map< string, float > > MedBootstrap::bootstrap	(	const MedFeatures &	features,
		map< int, map< string, map< string, float > > > *	results_per_split = NULL,
		with_registry_args *	registry_args = NULL
	)

Will run the bootstraping process on all cohorts and measurements.

MedFeatures need to contains also the information for the cohorts defenitions. for example: if there is Age:40-80, MedFeatures should contain Age Feature

Returns

the bootstrap results in map from cohort_name to all cohort measurements(a map). Each measurement is key,value in the map from measurement name to it's value if splits_inds is not NULL and mapping from each split value to it's coresponding indexes in the samples are provided - it will return also results for each split the higest level in the map is the split value

bootstrap() [2/4]

map< string, map< string, float > > MedBootstrap::bootstrap	(	MedSamples &	samples,
		const string &	rep_path,
		map< int, map< string, map< string, float > > > *	results_per_split = NULL,
		with_registry_args *	registry_args = NULL
	)

Will run the bootstraping process on all cohorts and measurements.

the input is samples, and rep_path. The rep_path is path to the repository which adds Age,Gender signals for creating the cohorts definitions. it's simple overload for convention

Returns

the bootstrap results in map from cohort_name to all cohort measurements(a map). Each measurement is key,value in the map from measurement name to it's value if splits_inds is not NULL and mapping from each split value to it's coresponding indexes in the samples are provided - it will return also results for each split the higest level in the map is the split value

bootstrap() [3/4]

map< string, map< string, float > > MedBootstrap::bootstrap	(	MedSamples &	samples,
		map< string, vector< float > > &	additional_info,
		map< int, map< string, map< string, float > > > *	results_per_split = NULL,
		with_registry_args *	registry_args = NULL
	)

Will run the bootstraping process on all cohorts and measurements.

the input is samples, additional_info. additional_info is provided for filtering and creating the cohorts. for example - Age:40-80 and Males

Returns

the bootstrap results in map from cohort_name to all cohort measurements(a map). Each measurement is key,value in the map from measurement name to it's value if splits_inds is not NULL and mapping from each split value to it's coresponding indexes in the samples are provided - it will return also results for each split the higest level in the map is the split value

bootstrap() [4/4]

map< string, map< string, float > > MedBootstrap::bootstrap	(	MedSamples &	samples,
		MedPidRepository &	rep,
		map< int, map< string, map< string, float > > > *	results_per_split = NULL,
		with_registry_args *	registry_args = NULL
	)

Will run the bootstraping process on all cohorts and measurements.

the input is samples, and rep. The rep is the repository which adds Age,Gender signals for creating the cohorts definitions. it's simple overload for convention

Returns

the bootstrap results in map from cohort_name to all cohort measurements(a map). Each measurement is key,value in the map from measurement name to it's value if splits_inds is not NULL and mapping from each split value to it's coresponding indexes in the samples are provided - it will return also results for each split the higest level in the map is the split value

change_sample_autosim() [1/2]

void MedBootstrap::change_sample_autosim	(	MedFeatures &	features,
		int	min_time,
		int	max_time,
		MedFeatures &	new_features
	)

changing the samples to be auto-simulations - taking max score in the time window for each pid

Returns

updates new_features from features

change_sample_autosim() [2/2]

void MedBootstrap::change_sample_autosim	(	MedSamples &	samples,
		int	min_time,
		int	max_time,
		MedSamples &	new_samples
	)

changing the samples to be auto-simulations - taking max score in the time window for each pid

Returns

updates new_samples from samples

filter_bootstrap_cohort() [1/3]

void MedBootstrap::filter_bootstrap_cohort ( const string &  rep, const string &  bt_json, MedSamples &  curr_samples, const string &  bt_cohort  )

static

commit bootstrap cohort filter on a given samples

Parameters

rep	- repository path
bt_json	- the json model to generate matrix for filtering the bootstrap cohort Automatically Age,Gender are added
curr_samples	- the samples to filter
bt_cohort	- a single line cohort (no support for MULTI) without the cohort name. only the filter definition. no tabs in the string.

Returns

filter samples from curr_samples by cohort definition

filter_bootstrap_cohort() [2/3]

void MedBootstrap::filter_bootstrap_cohort ( MedFeatures &  features, const string &  bt_cohort  )

static

commit bootstrap cohort filter on a given matrix

Parameters

features	- matrix
bt_cohort	- a single line cohort (no support for MULTI) without the cohort name. only the filter definition. no tabs in the string.

Returns

filter rows from features by cohort definition

filter_bootstrap_cohort() [3/3]

void MedBootstrap::filter_bootstrap_cohort ( MedPidRepository &  bt_repository, MedModel &  bt_filters, MedSamples &  curr_samples, const string &  bt_cohort  )

static

commit bootstrap cohort filter on a given samples

Parameters

bt_repository	- repository that was initialized for applying the bt_filters model to generate matrix
bt_filters	- the model to generate matrix for filtering the bootstrap cohort
curr_samples	- the samples to filter
bt_cohort	- a single line cohort (no support for MULTI) without the cohort name. only the filter definition. no tabs in the string.

Returns

filter samples from curr_samples by cohort definition

get_cohort_from_arg()

void MedBootstrap::get_cohort_from_arg

(

const string &

single_cohort

)

A function which reads a single cohort definition from the command line and parses it.

Please refer to parse_cohort_file for full spec.

init()

int MedBootstrap::init ( map< string, string > &  map )

virtual

Initialization string with format "parameter_name=value;..." each paramter_name is same as the class name field.

filter_cohort is path to file roc_Params is the init string for ROC_PARAMS

        if (param_name == "sample_ratio") {
            sample_ratio = stof(param_value);
            if (sample_ratio > 1.0 || sample_ratio < 0)
                MTHROW_AND_ERR("sample_ratio should be between 0-1, got %2.3f\n", sample_ratio);
        }
        else if (param_name == "sample_per_pid")
            sample_per_pid = stoi(param_value);
        else if (param_name == "loopcnt")
            loopCnt = stoi(param_value);
        else if (param_name == "sample_seed")
            sample_seed = stoi(param_value);
        else if (param_name == "sample_patient_label")
            sample_patient_label = stoi(param_value) > 0;
        else if (param_name == "roc_params")
            roc_Params = ROC_Params(param_value);
        else if (param_name == "filter_cohort") {
            filter_cohort.clear();
            parse_cohort_file(param_value);
        }
        else if (param_name == "simtimewindow")
            simTimeWindow = stoi(param_value) > 0;
        else if (param_name == "censor_time_factor")
            censor_time_factor = stof(param_value);
        else if (param_name == "is_binary_outcome")
            is_binary_outcome = stoi(param_value) > 0;
        else if (param_name == "use_time_control_as_case")
            use_time_control_as_case = stoi(param_value) > 0;
        else if (param_name == "sort_preds_in_multicategory")
            sort_preds_in_multicategory = stoi(param_value) > 0;
        else if (param_name == "measurement") {
            measurements_with_params.clear();
            //parse and create this object tuples need to pass in {}: format measure_name|param_init_string;...
            vector<string> tokens;
            boost::split(tokens, param_value, boost::is_any_of(";"));
            for (size_t i = 0; i < tokens.size(); ++i)
            {
                vector<string> parts;
                boost::split(parts, tokens[i], boost::is_any_of("|"));
                if (parts.size() != 2)
                    MTHROW_AND_ERR("Error MedBootstrap::init - expecting 2 tokens by | delimeter\n");
                MeasurmentFunctionType measure_type = measurement_function_name_to_type(parts[0]);
                MeasurementFunctions func = NULL;
                Measurement_Params *pr_m = NULL;
                string init_m_params = parts[1];
                switch (measure_type)
                {
                case MeasurmentFunctionType::calc_npos_nneg:
                    func = calc_npos_nneg;
                    break;
                case MeasurmentFunctionType::calc_only_auc:
                    func = calc_only_auc;
                    break;
                case MeasurmentFunctionType::calc_roc_measures_with_inc:
                    func = calc_roc_measures_with_inc;
                    pr_m = &roc_Params;
                    break;
                case MeasurmentFunctionType::calc_multi_class:
                    func = calc_multi_class;
                    pr_m = &multiclass_params;
                    multiclass_params.init_from_string(init_m_params);
                    is_binary_outcome = false;
                    break;
                case MeasurmentFunctionType::calc_kandel_tau:
                    func = calc_kandel_tau;
                    break;
                case MeasurmentFunctionType::calc_harrell_c_statistic:
                    func = calc_harrell_c_statistic;
                    break;
                case MeasurmentFunctionType::calc_regression:
                    func = calc_regression;
                    pr_m = &regression_params;
                    regression_params.init_from_string(init_m_params);
                    is_binary_outcome = false;
                    break;
                default:
                    MTHROW_AND_ERR("Error MedBootstrap::init unsupported measure %d\n", (int)measure_type);
                }
 
                measurements_with_params.push_back(pair<MeasurementFunctions, Measurement_Params*>(func, pr_m));
            }
 
        }

[MedBootstrap::init]

[MedBootstrap::init]

Reimplemented from SerializableObject.

measurement_function_name_to_type()

MeasurmentFunctionType MedBootstrap::measurement_function_name_to_type

(

const string &

measurement_function_name

)

convert measurement function name to type

Returns

MeasurmentFunctionType

parse_cohort_file()

void MedBootstrap::parse_cohort_file

(

const string &

cohorts_path

)

a function which reads cohorts file and stores it in filter_cohort.

The file format may be in 2 options:

1. COHORT_NAME[TAB]PARAMETERS_DEF - cohort name is string representing cohort
   name. PARAMETER_DEF is in format: "PARAMETER_NAME:MIN_RANGE,MAX_RANGE;..."
   the format can repeat itself with ";" between each parameter. the cohort
   will consist of intersection between all parameters ranges with "and" condition.
   there is single tab betwwen the name and the defenition.
   Example Line:
   1 year back & age 40-80 Time-Window:0,365;Age:40,80
   will create cohort called "1 year back & age 40-80" and will filter out records
   with (Time-Window>=0 and Time-Window<=365) and (Age>=40 and Age<=80)
   
2. MULTI[TAB]PARAMETERS_DEF[TAB]...PARAMETERS_DEF[TAB] - this definition with
   line starting with MULTI keyword will create all the cartesain options for each
   parameter definition with the each parameter definition in the next TABs.
   PARAMETERS_DEF - is same as option 1 format.
   Example Line:
   MULTI Time-Window:0,30;Time-Window:30,180 Age:40,60;Age:60,80;Age:40,80 Gender:1,1;Gender:2,2
   will create 2*3*2=12 cohorts for each Time-Window, Age, and Gender option
   

parse_cohort_line()

void MedBootstrap::parse_cohort_line

(

const string &

line

)

parsing specific line.

please refer to parse_cohort_file for full spec

prepare_bootstrap() [1/2]

void MedBootstrap::prepare_bootstrap	(	const MedFeatures &	features,
		vector< float > &	preds,
		vector< float > &	y,
		vector< int > &	pids,
		map< string, vector< float > > &	final_additional_info,
		vector< int > &	preds_order,
		unordered_map< int, vector< int > > *	splits_inds = NULL
	)

prepares the required vectors for bootstrap from MedFeatures &features

Returns

updates - preds, y, pids, final_additional_info with the information from MedFeatures. if splits_inds is provided (and not NULL) it will fill a mapping from split_index to the indexes in the samples vector correspond to each split value

prepare_bootstrap() [2/2]

void MedBootstrap::prepare_bootstrap	(	MedSamples &	samples,
		map< string, vector< float > > &	additional_info,
		vector< float > &	preds,
		vector< float > &	y,
		vector< int > &	pids,
		vector< int > &	preds_order,
		unordered_map< int, vector< int > > *	splits_inds = NULL
	)

prepares the required vectors for bootstrap from samples, additional_info

Returns

updates - preds, y, pids, final_additional_info with the information from samples, additional_info. if splits_inds is provided (and not NULL) it will fill a mapping from split_index to the indexes in the samples vector correspond to each split value

Field Documentation

measurement_function_name_map

unordered_map< string, MeasurmentFunctionType > MedBootstrap::measurement_function_name_map

static

Initial value:

= {
    { "calc_npos_nneg",MeasurmentFunctionType::calc_npos_nneg },
{ "calc_only_auc",MeasurmentFunctionType::calc_only_auc },
{ "calc_roc_measures_with_inc",MeasurmentFunctionType::calc_roc_measures_with_inc },
{ "calc_multi_class",MeasurmentFunctionType::calc_multi_class },
{ "calc_kandel_tau", MeasurmentFunctionType::calc_kandel_tau },
{ "calc_harrell_c_statistic", MeasurmentFunctionType::calc_harrell_c_statistic },
{ "calc_regression", MeasurmentFunctionType::calc_regression }
}

---

The documentation for this class was generated from the following files:

• Internal/MedStat/MedStat/MedBootstrap.h
• Internal/MedStat/MedStat/MedBootstrap.cpp