#include <gsl/gsl_permutation.h>
#include <gsl/gsl_sort_vector.h>
#include <gsl/gsl_statistics.h>
-#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "libpspp/assertion.h"
#include "libpspp/str.h"
#include "math/random.h"
-#include "output/tab.h"
+#include "output/pivot-table.h"
#include "output/text-item.h"
#include "gettext.h"
-static double
+static double
matrix_mindist (const gsl_matrix *m, int *mn, int *mm)
{
int i, j;
}
}
- return sqrt (mindist);
-}
-
-
-static void
-dump_matrix (const gsl_matrix *m)
-{
- size_t i, j;
-
- for (i = 0 ; i < m->size1; ++i)
- {
- for (j = 0 ; j < m->size2; ++j)
- printf ("%02f ", gsl_matrix_get (m, i, j));
- printf ("\n");
- }
+ return mindist;
}
const union value *val = case_data (c, qc->vars[j]);
if ( var_is_value_missing (qc->vars[j], val, qc->exclude))
NOT_REACHED ();
-
+
dist += pow2 (gsl_matrix_get (kmeans->centers, which, j) - val->f);
}
-
- return sqrt (dist);
+
+ return dist;
}
/* Return the minimum distance of the group WHICH and all other groups */
{
dist += pow2 (gsl_matrix_get (kmeans->centers, i, j) - gsl_matrix_get (kmeans->centers, which, j));
}
-
+
if (dist < mindist)
{
mindist = dist;
}
}
- return sqrt (mindist);
+ return mindist;
}
-/* Calculate the intial cluster centers. */
+/* Calculate the initial cluster centers. */
static void
kmeans_initial_centers (struct Kmeans *kmeans, const struct casereader *reader, const struct qc *qc)
{
}
}
else if (dist_from_case (kmeans, c, qc, mp) > min_dist_from (kmeans, qc, mq))
- /* If the distance between C and the second nearest group (MP) is greater than the
+ /* If the distance between C and the second nearest group (MP) is greater than the
smallest distance between the nearest group (MQ) and any other group, then replace
MQ with C */
{
dist += pow2 (gsl_matrix_get (kmeans->centers, i, j) - val->f);
}
- dist = sqrt (dist);
if (dist < mindist0)
{
if ( var_is_value_missing (qc->vars[j], val, qc->exclude))
missing = true;
}
-
+
if (missing)
continue;
double *x = gsl_matrix_ptr (kmeans->updated_centers, group, j);
*x += val->f * (qc->wv ? case_data (c, qc->wv)->f : 1.0);
}
- }
+ }
casereader_destroy (r);
}
*x /= n + 1; // Plus 1 for the initial centers
}
}
-
+
gsl_matrix_memcpy (kmeans->centers, kmeans->updated_centers);
{
kmeans->n = 0;
- int i;
/* Step 3 */
gsl_vector_long_set_all (kmeans->num_elements_groups, 0.0);
gsl_matrix_set_all (kmeans->updated_centers, 0.0);
struct ccase *c;
struct casereader *cs = casereader_clone (reader);
- for (; (c = casereader_read (cs)) != NULL; i++, case_unref (c))
+ for (; (c = casereader_read (cs)) != NULL; case_unref (c))
{
- int group = -1;
+ int group = -1;
kmeans_get_nearest_group (kmeans, c, qc, &group, NULL, NULL, NULL);
for (j = 0; j < qc->n_vars; ++j)
long *n = gsl_vector_long_ptr (kmeans->num_elements_groups, group);
*n += qc->wv ? case_data (c, qc->wv)->f : 1.0;
kmeans->n++;
-
-
}
casereader_destroy (cs);
static void
quick_cluster_show_centers (struct Kmeans *kmeans, bool initial, const struct qc *qc)
{
- struct tab_table *t;
- int nc, nr, currow;
- int i, j;
- nc = qc->ngroups + 1;
- nr = qc->n_vars + 4;
- t = tab_create (nc, nr);
- tab_headers (t, 0, nc - 1, 0, 1);
- currow = 0;
- if (!initial)
- {
- tab_title (t, _("Final Cluster Centers"));
- }
- else
- {
- tab_title (t, _("Initial Cluster Centers"));
- }
- tab_box (t, TAL_2, TAL_2, TAL_0, TAL_1, 0, 0, nc - 1, nr - 1);
- tab_joint_text (t, 1, 0, nc - 1, 0, TAB_CENTER, _("Cluster"));
- tab_hline (t, TAL_1, 1, nc - 1, 2);
- currow += 2;
-
- for (i = 0; i < qc->ngroups; i++)
- {
- tab_text_format (t, (i + 1), currow, TAB_CENTER, "%d", (i + 1));
- }
- currow++;
- tab_hline (t, TAL_1, 1, nc - 1, currow);
- currow++;
- for (i = 0; i < qc->n_vars; i++)
- {
- tab_text (t, 0, currow + i, TAB_LEFT,
- var_to_string (qc->vars[i]));
+ struct pivot_table *table = pivot_table_create (
+ initial ? N_("Initial Cluster Centers") : N_("Final Cluster Centers"));
+
+ struct pivot_dimension *clusters = pivot_dimension_create (
+ table, PIVOT_AXIS_COLUMN, N_("Cluster"));
+ clusters->root->show_label = true;
+ for (size_t i = 0; i < qc->ngroups; i++)
+ pivot_category_create_leaf (clusters->root,
+ pivot_value_new_integer (i + 1));
+
+ struct pivot_dimension *variables = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Variable"));
+ for (size_t i = 0; i < qc->n_vars; i++)
+ pivot_category_create_leaf (variables->root,
+ pivot_value_new_variable (qc->vars[i]));
+
+ const gsl_matrix *matrix = (initial
+ ? kmeans->initial_centers
+ : kmeans->centers);
+ for (size_t i = 0; i < qc->ngroups; i++)
+ for (size_t j = 0; j < qc->n_vars; j++)
+ {
+ double x = gsl_matrix_get (matrix, kmeans->group_order->data[i], j);
+ union value v = { .f = x };
+ pivot_table_put2 (table, i, j,
+ pivot_value_new_var_value (qc->vars[j], &v));
}
- for (i = 0; i < qc->ngroups; i++)
- {
- for (j = 0; j < qc->n_vars; j++)
- {
- if (!initial)
- {
- tab_double (t, i + 1, j + 4, TAB_CENTER,
- gsl_matrix_get (kmeans->centers,
- kmeans->group_order->data[i], j),
- var_get_print_format (qc->vars[j]), RC_OTHER);
- }
- else
- {
- tab_double (t, i + 1, j + 4, TAB_CENTER,
- gsl_matrix_get (kmeans->initial_centers,
- kmeans->group_order->data[i], j),
- var_get_print_format (qc->vars[j]), RC_OTHER);
- }
- }
- }
- tab_submit (t);
+ pivot_table_submit (table);
}
/* Reports cluster membership for each case. */
static void
quick_cluster_show_membership (struct Kmeans *kmeans, const struct casereader *reader, const struct qc *qc)
{
- struct tab_table *t;
- int nc, nr, i;
+ struct pivot_table *table = pivot_table_create (N_("Cluster Membership"));
- struct ccase *c;
- struct casereader *cs = casereader_clone (reader);
- nc = 2;
- nr = kmeans->n + 1;
- t = tab_create (nc, nr);
- tab_headers (t, 0, nc - 1, 0, 0);
- tab_title (t, _("Cluster Membership"));
- tab_text (t, 0, 0, TAB_CENTER, _("Case Number"));
- tab_text (t, 1, 0, TAB_CENTER, _("Cluster"));
- tab_box (t, TAL_2, TAL_2, TAL_0, TAL_1, 0, 0, nc - 1, nr - 1);
- tab_hline (t, TAL_1, 0, nc - 1, 1);
+ pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Cluster"),
+ N_("Cluster"));
+
+ struct pivot_dimension *cases = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Case Number"));
+ cases->root->show_label = true;
gsl_permutation *ip = gsl_permutation_alloc (qc->ngroups);
gsl_permutation_inverse (ip, kmeans->group_order);
- for (i = 0; (c = casereader_read (cs)) != NULL; i++, case_unref (c))
+ struct casereader *cs = casereader_clone (reader);
+ struct ccase *c;
+ for (int i = 0; (c = casereader_read (cs)) != NULL; i++, case_unref (c))
{
- int clust = -1;
assert (i < kmeans->n);
+ int clust;
kmeans_get_nearest_group (kmeans, c, qc, &clust, NULL, NULL, NULL);
- clust = ip->data[clust];
- tab_text_format (t, 0, i+1, TAB_CENTER, "%d", (i + 1));
- tab_text_format (t, 1, i+1, TAB_CENTER, "%d", (clust + 1));
+ int cluster = ip->data[clust];
+
+ int case_idx = pivot_category_create_leaf (
+ cases->root, pivot_value_new_integer (i + 1));
+ pivot_table_put2 (table, 0, case_idx,
+ pivot_value_new_integer (cluster + 1));
}
gsl_permutation_free (ip);
- assert (i == kmeans->n);
- tab_submit (t);
+ pivot_table_submit (table);
casereader_destroy (cs);
}
static void
quick_cluster_show_number_cases (struct Kmeans *kmeans, const struct qc *qc)
{
- struct tab_table *t;
- int nc, nr;
- int i, numelem;
- long int total;
- nc = 3;
- nr = qc->ngroups + 1;
- t = tab_create (nc, nr);
- tab_headers (t, 0, nc - 1, 0, 0);
- tab_title (t, _("Number of Cases in each Cluster"));
- tab_box (t, TAL_2, TAL_2, TAL_0, TAL_1, 0, 0, nc - 1, nr - 1);
- tab_text (t, 0, 0, TAB_LEFT, _("Cluster"));
-
- total = 0;
- for (i = 0; i < qc->ngroups; i++)
+ struct pivot_table *table = pivot_table_create (
+ N_("Number of Cases in each Cluster"));
+
+ pivot_dimension_create (table, PIVOT_AXIS_COLUMN, N_("Statistics"),
+ N_("Count"));
+
+ struct pivot_dimension *clusters = pivot_dimension_create (
+ table, PIVOT_AXIS_ROW, N_("Clusters"));
+ struct pivot_category *group = pivot_category_create_group (
+ clusters->root, N_("Cluster"));
+
+ long int total = 0;
+ for (int i = 0; i < qc->ngroups; i++)
{
- tab_text_format (t, 1, i, TAB_CENTER, "%d", (i + 1));
- numelem =
- kmeans->num_elements_groups->data[kmeans->group_order->data[i]];
- tab_text_format (t, 2, i, TAB_CENTER, "%d", numelem);
- total += numelem;
+ int cluster_idx = pivot_category_create_leaf (
+ group, pivot_value_new_integer (i + 1));
+ int count = kmeans->num_elements_groups->data[
+ kmeans->group_order->data[i]];
+ pivot_table_put2 (table, 0, cluster_idx,
+ pivot_value_new_integer (count));
+ total += count;
}
- tab_text (t, 0, qc->ngroups, TAB_LEFT, _("Valid"));
- tab_text_format (t, 2, qc->ngroups, TAB_LEFT, "%ld", total);
- tab_submit (t);
+ int cluster_idx = pivot_category_create_leaf (
+ clusters->root, pivot_value_new_text (N_("Valid")));
+ pivot_table_put2 (table, 0, cluster_idx, pivot_value_new_integer (total));
+ pivot_table_submit (table);
}
/* Reports. */
quick_cluster_show_results (struct Kmeans *kmeans, const struct casereader *reader, const struct qc *qc)
{
kmeans_order_groups (kmeans, qc); /* what does this do? */
-
+
if( qc->print_initial_clusters )
quick_cluster_show_centers (kmeans, true, qc);
quick_cluster_show_centers (kmeans, false, qc);
lex_error (lexer, NULL);
goto error;
}
- }
+ }
}
else if (lex_match_id (lexer, "PRINT"))
{
{
if (lex_match_id (lexer, "CLUSTERS"))
{
- if (lex_force_match (lexer, T_LPAREN))
+ if (lex_force_match (lexer, T_LPAREN) &&
+ lex_force_int (lexer))
{
- lex_force_int (lexer);
qc.ngroups = lex_integer (lexer);
if (qc.ngroups <= 0)
{
goto error;
}
lex_get (lexer);
- lex_force_match (lexer, T_RPAREN);
+ if (!lex_force_match (lexer, T_RPAREN))
+ goto error;
}
}
else if (lex_match_id (lexer, "CONVERGE"))
{
- if (lex_force_match (lexer, T_LPAREN))
+ if (lex_force_match (lexer, T_LPAREN) &&
+ lex_force_num (lexer))
{
- lex_force_num (lexer);
qc.epsilon = lex_number (lexer);
if (qc.epsilon <= 0)
{
goto error;
}
lex_get (lexer);
- lex_force_match (lexer, T_RPAREN);
+ if (!lex_force_match (lexer, T_RPAREN))
+ goto error;
}
}
else if (lex_match_id (lexer, "MXITER"))
{
- if (lex_force_match (lexer, T_LPAREN))
+ if (lex_force_match (lexer, T_LPAREN) &&
+ lex_force_int (lexer))
{
- lex_force_int (lexer);
qc.maxiter = lex_integer (lexer);
if (qc.maxiter <= 0)
{
goto error;
}
lex_get (lexer);
- lex_force_match (lexer, T_RPAREN);
+ if (!lex_force_match (lexer, T_RPAREN))
+ goto error;
}
}
else if (lex_match_id (lexer, "NOINITIAL"))
projects / pspp / blobdiff