/**
* Exports a distance matrix to a text file
* @param d Pointer to matrix
* @param fa Feature vector array
* @param file File name
*/
void export_dist(double *d, farray_t *fa, const char *file)
{
assert(d && fa && file);
int i, j;
FILE *f;
if (verbose > 0)
printf("Exporting distance matrix to '%s'.\n", file);
if (!(f = fopen(file, "w"))) {
error("Could not create file '%s'.", file);
return;
}
/* Print version header */
malheur_version(f);
/* Print distance header */
fprintf(f, "# ---\n# Distance matrix for %s\n", fa->src);
fprintf(f, "# Matrix size: %lu x %lu\n# ---\n", fa->len, fa->len);
fprintf(f, "# <report> <cluster> <dist1> <dist2> ... <distn>\n");
/* Print matrix */
for (i = 0; i < fa->len; i++) {
fprintf(f, "%s %s ", fa->x[i]->src, farray_get_label(fa, i));
for (j = 0; j < fa->len; j++)
fprintf(f, "%g ", d[i * fa->len + j]);
fprintf(f, "\n");
}
fclose(f);
}
/**
* Exports classification results
* @param p Prototype structure
* @param fa Feature vector array
* @param as Assignments to protoypes
* @param file File name
*/
void export_class(farray_t *p, farray_t *fa, assign_t *as, const char *file)
{
assert(p && fa && file);
int i, j;
char *l;
FILE *f;
if (verbose > 0)
printf("Exporting classification to '%s'.\n", file);
if (!(f = fopen(file, "w"))) {
error("Could not create file '%s'.", file);
return;
}
/* Print version header */
malheur_version(f);
/* Evaluate some quality functions */
double *e = quality(fa->y, as->label, as->len);
/* Print prototype header */
fprintf(f, "# ---\n# Classification for %s\n", fa->src);
fprintf(f, "# Precision of classification: %4.1f %%\n",
e[Q_PRECISION] * 100.0);
fprintf(f, "# Recall of classification: %4.1f %%\n",
e[Q_RECALL] * 100.0);
fprintf(f, "# F-measure of classification: %4.1f %%\n",
e[Q_FMEASURE] * 100.0);
fprintf(f, "# ---\n# <report> <label> <prototype> <distance>\n");
for (i = 0; i < fa->len; i++) {
j = as->proto[i];
l = as->label[i] ? farray_get_label(p, j) : "rejected";
fprintf(f, "%s %s %s %g\n", fa->x[i]->src, l, p->x[j]->src,
as->dist[i]);
}
fclose(f);
}
/**
* Exports a clustering structure to a text file
* @param c Clustering structure
* @param fa Feature vector array
* @param p Prototype struture
* @param a Assignments of prototypes
* @param file File name
*/
void export_cluster(cluster_t *c, farray_t *p, farray_t *fa, assign_t *a,
const char *file)
{
assert(c && fa && file);
FILE *f;
int i, j;
if (verbose > 0)
printf("Exporting clusters to '%s'.\n", file);
if (!(f = fopen(file, "w"))) {
error("Could not create file '%s'.", file);
return;
}
/* Print version header */
malheur_version(f);
/* Evaluate some quality functions */
double *e = quality(fa->y, c->cluster, c->len);
/* Print prototype header */
fprintf(f, "# ---\n# Clusters for %s\n", fa->src);
fprintf(f, "# Number of cluster: %lu\n", c->num);
fprintf(f, "# Precision of clusters: %4.1f %%\n",
e[Q_PRECISION] * 100.0);
fprintf(f, "# Recall of clusters: %4.1f %%\n", e[Q_RECALL] * 100.0);
fprintf(f, "# F-measure of clusters: %4.1f %%\n", e[Q_FMEASURE] * 100.0);
fprintf(f, "# ---\n# <report> <cluster> <prototype> <distance>\n");
for (i = 0; i < fa->len; i++) {
j = a->proto[i];
fprintf(f, "%s %s %s %g\n", fa->x[i]->src, cluster_get_name(c, i),
p->x[j]->src, a->dist[i]);
}
fclose(f);
}
/**
* Exports a structure of prototypes to a text file
* @param pr Prototype structure
* @param fa Feature vector array
* @param as Assignments to protoypes
* @param file File name
*/
void export_proto(farray_t *pr, farray_t *fa, assign_t *as, const char *file)
{
assert(pr && fa && file);
int i, j;
FILE *f;
if (verbose > 0)
printf("Exporting prototypes to '%s'.\n", file);
if (!(f = fopen(file, "w"))) {
error("Could not create file '%s'.", file);
return;
}
/* Print version header */
malheur_version(f);
/* Evaluate some quality functions */
double *e = quality(fa->y, as->proto, as->len);
/* Print prototype header */
fprintf(f, "# ---\n# Prototypes for %s\n", fa->src);
fprintf(f, "# Number of prototypes: %lu\n", pr->len);
fprintf(f, "# Compression of prototypes: %4.1f %%\n",
pr->len * 100.0 / (double) fa->len);
fprintf(f, "# Precision of prototypes: %4.1f %%\n",
e[Q_PRECISION] * 100.0);
fprintf(f, "# ---\n# <report> <prototype> <distance>\n");
for (i = 0; i < fa->len; i++) {
j = as->proto[i];
fprintf(f, "%s %s %g\n", fa->x[i]->src, pr->x[j]->src, as->dist[i]);
}
fclose(f);
}
/**
* Exports results from the incremental analysis (phase 1). The results are
* obtained by first classifying and then clustering reports. In the first
* phase the classified reports are written to the output file.
* @param p Prototype struture
* @param fa Feature vector array
* @param as Assignments to prototypes
* @param file File name
*/
void export_increment1(farray_t *p, farray_t *fa, assign_t *as,
const char *file)
{
int i, j;
FILE *f;
if (verbose > 0)
printf("Exporting incremental analysis (1) to '%s'.\n", file);
if (!(f = fopen(file, "w"))) {
error("Could not create file '%s'.", file);
return;
}
/* Print version header */
malheur_version(f);
/* Print incremental header */
fprintf(f, "# ---\n# Incremental analysis for %s\n", fa->src);
fprintf(f, "# ---\n# <report> <cluster> <prototype> <distance>\n");
if (!p || !as) {
fclose(f);
return;
}
for (i = 0; i < fa->len; i++) {
if (!as->label[i])
continue;
j = as->proto[i];
fprintf(f, "%s %s %s %g\n", fa->x[i]->src, farray_get_label(p, j),
p->x[j]->src, as->dist[i]);
}
fclose(f);
}
/**
* Parse command line options
* @param argc Number of arguments
* @param argv Argument values
*/
static void parse_options(int argc, char **argv)
{
int ch;
/* reset getopt */
optind = 0;
while ((ch = getopt_long(argc, argv, OPTSTRING, longopts, NULL)) != -1) {
switch (ch) {
case 'n':
save = FALSE;
break;
case 'r':
reset = TRUE;
break;
case 'v':
case 'm':
/* Empty. See load_config() */
break;
case 'o':
output_file = optarg;
break;
case 'V':
malheur_version(stdout);
exit(EXIT_SUCCESS);
break;
case 'h':
case '?':
print_usage();
exit(EXIT_SUCCESS);
break;
/* long options */
case 1001:
config_set_string(&cfg, "input.format", optarg);
break;
case 1002:
config_set_int(&cfg, "input.mist_level", atoi(optarg));
break;
case 1003:
config_set_int(&cfg, "input.mist_rlen", atoi(optarg));
break;
case 1004:
config_set_int(&cfg, "input.mist_tlen", atoi(optarg));
break;
case 1005:
config_set_string(&cfg, "features.ngram_delim", optarg);
break;
case 1006:
config_set_int(&cfg, "features.ngram_len", atoi(optarg));
break;
case 1007:
config_set_string(&cfg, "features.vect_embed", optarg);
break;
case 1008:
config_set_int(&cfg, "features.lookup_table", atoi(optarg));
break;
case 1009:
config_set_float(&cfg, "prototypes.max_dist", atof(optarg));
break;
case 1010:
config_set_int(&cfg, "prototypes.max_num", atoi(optarg));
break;
case 1011:
config_set_float(&cfg, "classify.max_dist", atof(optarg));
break;
case 1012:
config_set_string(&cfg, "cluster.link_mode", optarg);
break;
case 1013:
config_set_float(&cfg, "cluster.min_dist", atof(optarg));
break;
case 1014:
config_set_int(&cfg, "cluster.reject_num", atoi(optarg));
break;
case 1015:
config_set_int(&cfg, "cluster.shared_ngrams", atoi(optarg));
break;
}
}
/* Check configuration */
config_check(&cfg);
argc -= optind;
argv += optind;
if (argc < 1)
fatal("the <action> argument is required");
/* Argument: action */
if (!strcasecmp(argv[0], "prototype")) {
action = PROTOTYPE;
} else if (!strcasecmp(argv[0], "distance")) {
action = DISTANCE;
} else if (!strcasecmp(argv[0], "cluster")) {
action = CLUSTER;
} else if (!strcasecmp(argv[0], "classify")) {
action = CLASSIFY;
} else if (!strcasecmp(argv[0], "increment")) {
action = INCREMENT;
} else if (!strcasecmp(argv[0], "protodist")) {
action = PROTODIST;
} else if (!strcasecmp(argv[0], "info")) {
action = INFO;
} else {
fatal("Unknown analysis action '%s'", argv[0]);
}
if (argc < 2 && action != PROTODIST && action != INFO)
fatal("the <dataset> argument is required");
/* Assign input files */
input_files = argv + 1;
input_len = argc - 1;
}
