TkbAnalytic               package:FKBL               R Documentation

_A_n_a_l_y_t_i_c _a_l_g_o_r_i_t_h_m. _T_w_e_a_k_s _a _k_B

_D_e_s_c_r_i_p_t_i_o_n:

     This algorithm is called Analytic.  It is the same as
     TkbAnalyticL, but it has not the parameters packed in a list. It
     takes a knowledge base and  tweaks its weights to fit better a
     given train data. It is based in the idea of checking every
     training case, and for the cases whose class wasn't correctly
     inferred, it tries to improve the rule set. This improvement is
     done by examining three different solutions. First, lowering the
     winning rule's weight so it is no longer the winning rule in this
     situation. Second, find a non winning rule which wouldn't have
     made a mistake, and rise its weight up to the value which makes it
     to win. Third, leave it as is. The situation which gives less
     error with the whole training set, would be the chosen one. This
     algorithm is described in chapter 3, pages 048-057 at Ishibuchi et
     al.$

_U_s_a_g_e:

      TkbAnalytic(kB, itera, e, train)

_A_r_g_u_m_e_n_t_s:

     Takes knowledge base, the maximum number of iterations, the e and
     the train  data.

      kB: The knowledge base to tweak.

   itera: The maximum number of iterations.

       e: The e parameter for analytic.

   train: The train data.

_V_a_l_u_e:

     Returns the tweaked knowledge base.

_S_o_u_r_c_e:

     \begin{itemize}

     *  Ishibuchi, H., Nakashima, T., Nii, M.

     *  "Classification and modeling with linguistic information
        granules." 

     *  Soft Computing Approaches to Linguistic Data Mining. 

     *  Springer-Verlag, 2003 \end{itemize}

_E_x_a_m_p_l_e_s:

      data(kB)
      data(trainA)
      TkbAnalytic(kB, 1000, 0.1, trainA)

