clc.c

聚类算法全集以及内附数据集

    #endif
    printf("%s\n", DESCRIPTION);
    printf("%s\n", VERSION);
    printf("-m#      conjunction for membership degrees         "
                    "(default: product)\n");
    printf("-c#      conjunction for coincidence matrix entries "
                    "(default: product)\n");
    printf("-b/f/r#  blank characters, field and record separators\n"
           "         (default: \" \\t\\r\", \" \\t\", \"\\n\")\n");
    printf("clsfile  file to read cluster set description from\n");
    #ifdef MATVERSION
    printf("patfile  pattern file to read (no header, only numbers)\n");
    #else
    printf("-d       use default header "
                    "(field names = field numbers)\n");
    printf("-h       read table header (field names) from hdrfile\n");
    printf("hdrfile  file containing table header (field names)\n");
    printf("tabfile  table file to read "
                    "(field names in first record)\n");
    #endif
    return 0;                   /* print a usage message */
  }                             /* and abort the program */

  /* --- evaluate arguments --- */
  for (i = 1; i < argc; i++) {  /* traverse arguments */
    s = argv[i];                /* get option argument */
    if (optarg) { *optarg = s; optarg = NULL; continue; }
    if ((*s == '-') && *++s) {  /* -- if argument is an option */
      while (1) {               /* traverse characters */
        switch (*s++) {         /* evaluate option */
          case '!': help();                break;
          case 'b': optarg = &blanks;      break;
          case 'f': optarg = &fldseps;     break;
          case 'r': optarg = &recseps;     break;
          case 'm': optarg = &conjfn1;     break;
          case 'c': optarg = &conjfn2;     break;
          #ifndef MATVERSION
          case 'd': flags |= AS_DFLT;      break;
          case 'h': optarg = &fn_hdr;      break;
          #endif
          default : error(E_OPTION, *--s); break;
        }                       /* set option variables */
        if (!*s) break;         /* if at end of string, abort loop */
        if (optarg) { *optarg = s; optarg = NULL; break; }
      } }                       /* get option argument */
    else {                      /* -- if argument is no option */
      switch (k++) {            /* evaluate non-option */
        case  0: fn_cls1 = s;     break;
        case  1: fn_cls2 = s;     break;
        case  2: fn_in   = s;     break;
        default: error(E_ARGCNT); break;
      }                         /* note filenames */
    }
  }
  if (optarg) error(E_OPTARG);  /* check option argument and */
  if (k != 3) error(E_ARGCNT);  /* the number of arguments */
  i = (!fn_cls1 || !*fn_cls1) ? 1 : 0;
  if  (!fn_cls2 || !*fn_cls2) i++;
  if  (!fn_in   || !*fn_in  ) i++;
  #ifndef MATVERSION
  if (fn_hdr) {                 /* set the header file flag */
    flags = AS_ATT | (flags & ~AS_DFLT);
    if (strcmp(fn_hdr, "-") == 0) fn_hdr = "";
    if (!*fn_hdr) i++;          /* convert "-" to "" and */
  }                             /* count assignment of stdin */
  #endif                        /* check assignments of stdin: */
  if (i > 1) error(E_STDIN);    /* stdin must not be used twice */
  cfc1 = code(conjtab,conjfn1); /* get and check first  conjunction */
  if (cfc1 < 0) error(E_CONJFN, conjfn1);
  cfc2 = code(conjtab,conjfn2); /* get and check second conjunction */
  if (cfc2 < 0) error(E_CONJFN, conjfn2);

  /* --- read first cluster set --- */
  scan = sc_create(fn_cls1);     /* create a scanner */
  if (!scan) error((!fn_cls1 || !*fn_cls1) ? E_NOMEM:E_FOPEN, fn_cls1);
  fprintf(stderr, "\nreading %s ... ", sc_fname(scan));
  if (sc_nexter(scan) < 0) error(E_PARSE, sc_fname(scan));
  #ifdef MATVERSION
  clset1 = cls_parse(scan);     /* parse the input cluster set */
  #else
  attset = as_create("domains", att_delete);
  if (!attset) error(E_NOMEM);  /* create an attribute set */
  if ((as_parse(attset, scan, AT_ALL) != 0)
  ||  (as_attcnt(attset) <= 0)) /* parse the attribute set */
    error(E_PARSE, sc_fname(scan));
  clset1 = cls_parsex(scan, attset, 0);
  #endif                        /* parse the cluster set */
  if (!clset1 || !sc_eof(scan)) error(E_PARSE, sc_fname(scan));
  #ifdef MATVERSION
  attcnt = cls_incnt(clset1);   /* get the number of attributes */
  #else
  attcnt = as_attcnt(attset);   /* get the number of attributes */
  #endif
  clscnt = cls_clscnt(clset1);  /* get the number of clusters */
  fprintf(stderr, "[%d attribute(s), ",   attcnt);
  fprintf(stderr, "%d cluster(s)] done.", clscnt);
  sc_delete(scan); scan = NULL; /* delete the scanner */

  /* --- read second cluster set --- */
  scan = sc_create(fn_cls2);     /* create a scanner */
  if (!scan) error((!fn_cls2 || !*fn_cls2) ? E_NOMEM:E_FOPEN, fn_cls1);
  fprintf(stderr, "\nreading %s ... ", sc_fname(scan));
  if (sc_nexter(scan) < 0) error(E_PARSE, sc_fname(scan));
  #ifdef MATVERSION
  clset2 = cls_parse(scan);     /* parse the input cluster set */
  #else
  dummy = as_create("domains", att_delete);
  if (!dummy) error(E_NOMEM);   /* create an attribute set */
  if ((as_parse(dummy, scan, AT_ALL) != 0)
  ||  (as_attcnt(dummy) <= 0))  /* parse the attribute set */
    error(E_PARSE, sc_fname(scan));
  as_delete(dummy);             /* delete the attribute set */
  clset2 = cls_parsex(scan, attset, 0);
  #endif                        /* parse the cluster set */
  if (!clset2 || !sc_eof(scan)) error(E_PARSE, sc_fname(scan));
  i = cls_clscnt(clset2);       /* get the number of clusters */
  fprintf(stderr, "[%d attribute(s), ",   attcnt);
  fprintf(stderr, "%d cluster(s)] done.", i);
  sc_delete(scan); scan = NULL; /* delete the scanner */
  if (i != clscnt) error(E_DIFFCNT, clscnt, i);

  /* --- create partition matrices --- */
  pmat1 = pm_create(clscnt);    /* create a partition matrix */
  pmat2 = pm_create(clscnt);    /* for each of the cluster sets */
  if (!pmat1 && !pmat2) error(E_NOMEM);

  #ifdef MATVERSION
  /* --- process patterns --- */
  if (fn_in && *fn_in)          /* if an file name is given, */
    in = fopen(fn_in, "r");     /* open the file for reading */
  else {                        /* if no fine name is given, */
    in = stdin; fn_in = "<stdin>"; }       /* use std. input */
  fprintf(stderr, "\nreading %s ... ", fn_in);
  if (!in) error(E_FOPEN, fn_in);
  tfscan = tfs_create();        /* create a table file scanner and */
  if (!tfscan) error(E_NOMEM);  /* set the separator characters */
  if (blanks)  tfs_chars(tfscan, TFS_BLANK,  blanks);
  if (fldseps) tfs_chars(tfscan, TFS_FLDSEP, fldseps);
  if (recseps) tfs_chars(tfscan, TFS_RECSEP, recseps);
  err = tfs_err(tfscan);        /* get the error information */
  pat = vec_readx(tfscan, in, &attcnt);
  if (!pat) {                   /* read the first training pattern */
    if (err->code >= 0) error(E_FREAD, fn_in);
    error(err->code, fn_in, 1, err->s, err->fld, err->exp);
  }                             /* check for success */
  do {                          /* pattern read loop */
    cls_exec(clset1, pat,NULL); /* execute the first cluster set */
    if (pm_addcol(pmat1, clset1) < 0)
      error(E_NOMEM);           /* store the membership degrees */
    cls_exec(clset2, pat,NULL); /* execute the second cluster set */
    if (pm_addcol(pmat2, clset2) < 0)
      error(E_NOMEM);           /* store the membership degrees */
    patcnt++;                   /* count pattern and read next */
  } while (vec_read(pat, attcnt, tfscan, in) == 0);
  if (err->code < 0)            /* check for an error */
    error(err->code, fn_in, patcnt +1, err->s, err->fld, err->exp);
  if (tfs_delim(tfscan) != TFS_EOF)  /* check for end of file */
    error(E_VALUE, fn_in, patcnt +1, "\"\"", 1);
  if (in != stdin) fclose(in);  /* close the input file and */
  in = NULL;                    /* clear the file variable */
  fprintf(stderr, "[%d pattern(s)] done.\n", patcnt);

  #else
  /* --- process patterns --- */
  fprintf(stderr, "\n");        /* terminate previous log message */
  as_chars(attset, blanks, fldseps, recseps, "");
  in = io_hdr(attset, fn_hdr, fn_in, flags|AS_NOXATT, 1);
  if (!in) error(1);            /* read the table header */
  k = AS_INST | (k & ~AS_ATT);  /* write the attribute names */
  f = AS_INST | (flags & ~(AS_ATT|AS_DFLT));
  i = ((flags & AS_DFLT) && !(flags & AS_ATT))
    ? 0 : as_read(attset, in, f);
  while (i == 0) {              /* record read loop */
    tplcnt++;                   /* count tuple and sum its weight */
    tplwgt += w = as_getwgt(attset);
    cls_valuex(clset1, NULL);   /* set the data from a tuple and */
    cls_exec(clset1,NULL,NULL); /* execute the first cluster set */
    if (pm_addcol(pmat1, clset1) < 0)
      error(E_NOMEM);           /* store the membership degrees */
    cls_valuex(clset2, NULL);   /* set the data from a tuple and */
    cls_exec(clset2,NULL,NULL); /* execute the second cluster set */
    if (pm_addcol(pmat2, clset2) < 0)
      error(E_NOMEM);           /* store the membership degrees */
    i = as_read(attset, in, f); /* try to read the next record */
  }
  if (i < 0) {                  /* if an error occurred, */
    err = as_err(attset);       /* get the error information */
    tplcnt += (flags & (AS_ATT|AS_DFLT)) ? 1 : 2;
    io_error(i, fn_in, tplcnt, err->s, err->fld, err->exp);
    error(1);                   /* print an error message */
  }                             /* and abort the program */
  if (in != stdin) fclose(in);  /* close the table file and */
  in = NULL;                    /* clear the file variable */
  fprintf(stderr, "[%d/%g tuple(s)] done.\n", tplcnt, tplwgt);
  #endif

  /* --- compute and print evaluations --- */
  for (i = 8; --i >= 0; ) eval[i] = 0;
  if (pm_cmp(pmat1, pmat2, cfc1, cfc2, eval) < 0)
    error(E_NOMEM);             /* compare the partition matrices */
  printf("mean squared difference      : %g\n", eval[PM_DIFF]);
  printf("cross-classification accuracy: %g\n", eval[PM_ACC]);
  printf("F1 measure                   : %g\n", eval[PM_F1]);
  printf("Rand statistic               : %g\n", eval[PM_RAND]);
  printf("Jaccard coefficient          : %g\n", eval[PM_JACCARD]);
  printf("Folkes-Mallows index         : %g\n", eval[PM_FOLKES]);
  printf("Hubert index                 : %g\n", eval[PM_HUBERT]);

  /* --- clean up --- */
  #ifndef NDEBUG
  #ifdef MATVERSION
  tfs_delete(tfscan);           /* delete the table file scanner */
  #else
  as_delete(attset);            /* delete the attribute set */
  #endif
  pm_delete(pmat1);             /* delete the partition matrices */
  pm_delete(pmat2);
  cls_delete(clset1);           /* delete the cluster sets */
  cls_delete(clset2);
  #endif
  return 0;                     /* return 'ok' */
}  /* main() */