Page 254 - Contributed Paper Session (CPS) - Volume 3

CPS2011 Dominique H. et al.
current weight vector (): ( + 1) = () + ℎ ()[() − ()] ,

where ℎ () is a neighbourhood function, represents the winning node.

The simplest neighbourhood function is a monotonically decreasing
−(,)
Gaussian function: ℎ () = ∗ exp( 2 () ) , where represents the

2
learning rate, and () is the width of the kernel.

Figure 1: SOM with 2 - D Input, 3 x 3 Output, obtamed from “DATA MINING
Concepts, Models, Methods, and Algorithms”

3. Results
The data were sourced from Bloomberg for the 67 consumer discretionary
firms listed on the S&P 500 for the years 2006-2016. During the data pre-
processing stage, it was decided to consider 27 firms with 2006-16 as the time-
window and 19 financial ratios since it is prudent to not impute missing values
in a financial time-series, which can display unusual jumps. The 11-year period
covers at least one full economic cycle and encompasses the sub-prime
period.
The first-step involved building a confirmatory factor analysis (CFA) model.
Nineteen indicators were used initially and were pruned to thirteen listed in
Table 1 after removing variables during an iterative procedure. The final CFA
model is displayed in Figure 2. The SmartPLS 2.0 package was used to carry
out this analysis.
As shown in Figure 2, we observe a positive association between liquidity
and profitability, and we see that profitability decreases as activity or debt
increases. Eventually, profitability positively contributes to the market factor.

243 | I S I W S C 2 0 1 9