A Framework for Bus Trajectory Extraction and Missing Data Recovery for Data Sampled from the Internet

. 2017 Feb 10;17(2):342. doi: 10.3390/s17020342

Algorithm 2 FCM clustering algorithm.

Input:

G^{(m, l)}

, N (the number of data in

G^{(m, l)}

, i.e.,

N = | G^{(m, l)} |

);

Output: U (the partition matrix),

G_{c}^{(m, l)}

(the aggregated clusters);

Process:

(a1) get $U^{F - F C M}$ , $V^{F - F C M}$ , $J_{a}^{F - F C M}$ by F-FCM clustering;
(a2) get $U^{B - F C M}$ , $V^{B - F C M}$ , $J_{a}^{B - F C M}$ by B-FCM clustering;
(a3) clustering method determination:
if $J_{a}^{F - F C M} \leq J_{a}^{B - F C M}$ then
set $U = U^{F - F C M}$ ;
else
set $U = U^{B - F C M}$ ;
end if
(a4) partition loop:
for i← 1 to N do
if $u_{i j} = \max_{k = 1, 2, \dots, c} u_{i k}$ then
partition $G_{i}^{(m, l)}$ (the i-th data in $G^{(m, l)}$ ) into $G_{c}^{(j, m, l)}$ ;
end if
end for
(a5) generate $G_{c}^{(m, l)}$ by $G_{c}^{(m, l)} = \cup_{j = 1}^{c} G_{c}^{(j, m, l)}$ .