Fuzzy Clustering-based Model for Productivity Forecasting

madbrainedmudlickΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

78 εμφανίσεις

Fuzzy Clustering
-
based Model

for Productivity
Forecasting



Seyed Farid Mirahadi
1

Tarek Zayed
2

Emad Elwakil
3


1
Graduate Student, Department of Building, Civil & Environmental Engineering, Concordia
University, E
-
mail: S_miraha@encs.concordia.ca

2
Associate

Professor
, Department of Building, Civil & Environmental Engineering, Concordia
University, E
-
mail: Zayed@encs.concordia.ca

3
Assistant Professor, Department of Civil Engineering and Construction Management, California
State
University,

Northridge
, E
-
mail: Elwakil@csun.edu


Abstract

Forecasting

productivity of

construction
operations

is a difficult but
crucial

task in
planning construction projects
. Over the past decades, many models have been developed to
forecast produ
ctivity for different construction operations. Models made up of several functional
relations and controlled by a specific number of control rules are more in line with human
reasoning and logic. Such models can determine which relation to follow. Moreover
, with the
increasing heterogeneity of model's data, the difficulty of building an efficient structure with only
one comprehensive relation increases. This matter instigated the idea of using clustered data
space. A
clustering
-
based neural network

model ba
sed on integration of
clustering techniques
,
artificial neural network
(ANN)

and
fuzzy reasoning

is one way to model the clustered data
space. This research presents a framework that improves the accuracy of productivity
forecasting models. The main contribution of the proposed model is the employment of a
clustering method to acquire sub
-
sets of tra
ining data prior to learning phase.
Fuzzy C
-
Means
is
applied to select meaningful sub
-
sets that train each neural network independently. Meanwhile,
a distinct control neural network, which estimates the degree of membership
s

to determined
clusters, helps i
n selecting the best sub
-
model that fits every set of testing data. The proposed
model is further verified through simulation
in which several qualitative and quantitative factors
are considered. It is implemented to a case study, which shows a considerabl
e improvement of
model performance with lower
Mean Squared Error
.

The developed
model

assists researchers
and practitioners in utilizing historical construction data to
forecast
productivity
of construction
operations
that could not be obtained by traditio
nal techniques
.




Keywords:

productivity forecasting; fuzzy reasoning; fuzzy clustering; neural network; clustering
-
based
model