B) that specifies the input weights randomly and the

B)      Extreme learning machine (ELM)

An
extreme learning machine (ELM) is a type of NN that is used (Equation 8) for a
single-layer feed-forward NN that specifies the input weights randomly and the
output weights analytically 41, 42.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

In the
following equation,

 are
learning parameters of activation functions and

 is the weight
of the kth hidden neuron associated to the output neuron. The input
data are

, and N is the number of samples where

.

 (8)

So:

(9)

=g(

)                                 (10)

=g(

)                             (11)

         and

(12)

 

1. Discussion and Experimental results

Zi?ba et
al. performed boosted SVM to discover rules from a real clinical data set for post-operative
survival expectancy in lung cancer patients, and they extracted 16 features for
predicting post-operative survival expectancy. The data set was gathered retroactively
for lung samples from 1200 lung cancer patients in the years 2007–2011 at the Wroclaw
thoracic surgery station, which was managed by the Institute of Tuberculosis and
Pulmonary Diseases in Warsaw, Poland 21.

The main data set contained 139 features, of which
36 were from pre-operative, 37 from peri-operative and 46 (containing 17 pathology-related)
from post-operative periods.

Zi?ba et al. used 36 pre-operative features for
the prediction of 1-year survival, and they reduced the number of features from
36 to 16 with 470 example records by the boosted SVM method.

Iraji predicted
1-year post-operative
survival expectancy for thoracic lung cancer surgery by applying MLA-ANFIS, neural networks, regression and ELM,
based on the same thoracic surgery data set with 16 input features 22.

We tried
to design a careful method for predicting post-operative survival expectancy in
lung cancer patients with a thoracic surgery data set 21; 470 patient records
were used, and the 16 input variable features are specified below.

 

v1. DGN: Diagnosis – specific combination of ICD

v2. PRE5: Volume exhaled at the end of the first
second of forced expiration – FEV1 (numeric)

v3. PRE4: Forced vital capacity – FVC

v4. PRE7: Pain before surgery

v5. PRE6: Performance status – Zubrod scale

v6. PRE8: Haemoptysis before surgery

v7. PRE10: Cough before surgery

v8. PRE11: Weakness before surgery

v9. PRE9: Dyspnoea before surgery

v10. PRE14: T in clinical TNM – size of the original
tumour

v11. PRE17: Type 2 DM – diabetes mellitus

v12. PRE19: MI up to 6 months

v13. PRE25: PAD – peripheral arterial diseases

v14. PRE30: Smoking

v15. PRE32: Asthma

v16. AGE: Age at surgery

Output: Risk1Y: True value if died in 1-year survival
period

 

Of the patients with lung cancer disease in the
data set, 400 patients survived (positive) and 70 patients died (negative); the
number of patients who survived after surgery is greater than the number of patients
who died. Choosing patients to undergo lung cancer surgery who have a lower risk
of death during the short term (30-day period) or long term (1- or 5-year period)
after surgery is very important 21.

In this paper, we evaluated 1-year survival, to
estimate post-operative survival expectancy in lung cancer patients.

We executed
our proposed method in MATLAB version 9.1.0.441655 (R2016b) on a laptop with 1.7
GHz CPU, and we applied root mean square error (RMSE) and
accuracy in order to calculate evaluation
indices, to compare our method with others and to identify the best one. RMSE
describes the sample standard deviation of the
differences between predicted values and actual values, and accuracy
demonstrates precision classification.