Behavioral based risk prediction-1

1.INTRODUCTION

One important goal of any field of science is to develop a theory (or model) that predicts future outcomes. .Risk prediction and the usage of such prediction for the improvement of safety at the workplace would be a gift for the most managers.

Risk could be considered from a managerial point of view a combination between the prediction (and mitigation) of eventual unpleasant events and the reward that could be gained  upon competitors going with a certain risk.

The paper presents some research done by INCDPM regarding occupational risk predicitve analysis coupled with predicitve training. Having available the necessary statistical data, prediction could be done and could be improved step by step in order to come as close as possible to reality. The introduction of loss function as a referential criteria in prediction could help moderate the prediction done in order to use as much as possible such tool in risk mitigation and also improvement of performance in the spirit of ISO 31.000/2011.

 

2.METHODS USED IN THE RESEARCH

In order to obtain the best solution, as research methods were chosen beyond the specific methods from risk analysis also more general managerial methods like 8D and Six Sigma.The managerial approach is used mainly in order to improve the risk assessment process and also to optimize the predictive training.

Prediction is based upon a Bayesian approach and also upon neural network. Some  of the most interesting aspects are described below.

In order that the prediction should be succesfull  there are used models of enterprises. Such a model could be borrowed from allready defined models (for example for process services like maintenance) or could be developed from scratch using specific tools- like PROMIS.

 

 

2.1.THE BAYESIAN APPROACH

In the predictive Bayesian approach there are no fictional parameters introduced, and references to statistic based probabilities. The focus is on the observable quantities of risk, and the actual population exăposed to this risk. The observable quantities are quantities that express states of the "world of risk"; quantities that are unknown at the time of the analysis but will, if the system/activity actually is implemented, take some value in the future and possibly become known.Applying the predictive approach, it is possible, for example to focus on the number of Unpredicted (accidental) events X in some specified work operations. By modelling (for example using event trees and fault trees) it could be  established  a link g between X and observables on a more detailed system level, denoted Z = (Z1, Z2, ..., Zm). Here Zi could be the number of risk prone situations of a certain type occurring during some operations or an indicator function which equals 1 if a specific safety barrier fails and 0 otherwise.

As an example, it could be considered a prediction of X by the following reasoning: General statistical data show that the falling hazard typical occurs 20 times, and hence the prediction Z1* = 20 of Z1. To assess the first safety barrier performance-which is training, consider 10 hazardous situations. Say that the system predicts 1 failure. Then this gives aprobability P(Z2 = 1) = 0.1. Now given that the hazardous situation occurs and the first safety barrier fails, how reliable is the second safety barrier? Say that we assign a probability of 0.4 in this case. That means that we predict 4 failures out of 10 cases. Or we could simply say that 0.4 reflects the assessor's uncertainty. Hence P(Z2 = 1 | Z1 = 1) = 0.4, and using the model g, we obtain the prediction X* = Z1*·Z2*·Z3* = 20·0.1·0.4 = 0.8.

The prediction is  1 accidental event for a manual handling operation which involves the possibility of a fall accident.. Next there is a need to address the associated uncertainties. One way of doing this is to specify a 90% prediction interval for Z1. If the interval is [10, 100], the system  is 90% certain that the number of hazardous situations Z1 will be in this interval, given the background information of the analysis. The interval is determined based on the available information, that is, relevant data and expert judgments. From this analysis results a 90% prediction interval for X equal to [0, 4], using the same probabilities for the barrier safety failures as above.Furthermore, the assumptions and suppositions of the probability assignments provide an additional checklist Some special consequences could be underlined, as mentioned below:

a) Delay effects – which describe the time of latency between an initial event and the actual damage.

b) Reversibility – which describe the possibility – using training- to restore the situation to the state before damage occurred. This feature classification system can be used as a checklist for ensuring the right focus of the analysis, and  it can also be used as a checklist for identifying relevant uncertainty factors. For example, the feature "delay effects" could lead to a focus on activities or mechanisms that could initiate deteriorating processes causing predictible events that could be supressed through training. Addressing the uncertainties also mean to consider the manageability; i.e. to what extent is it possible to control and reduce the uncertainties, and obtain desirable outcomes? Some risks are more manageable than others, meaning that the potential for reducing the risk is larger for some risks compared to others. By proper uncertainty management, it is possible to obtain desirable consequences. Expressing risk, also means to perform sensitivity analyses. The purpose of these analyses is to show how sensitive the output risk indices are with respect to changes in basic input quantities, for example assumptions and suppositions. Risk is described by addressing such issues along with the probabilities. It gives a sound basis for risk analysis in general and for predictive training in particular.

2.2.DEVELOPING A NEURAL NETWORK RISK PREDICTION TOOL

 

The main steps are:

1.      Collect data

2.      Create the network

3.      Configure the network

4.      Initialize the weights and biases

5.      Train the network

6.      Validate the network

7.      Use the network

 

To define a  risk pattern recognition problem it is possible to  arrange a set of Q input vectors as columns in a matrix. Then arrange another set of Q target vectors so that they indicate the risk classes to which the input vectors are assigned . There are two approaches to creating the target vectors.

The most simple  approach can be used when there are only two classes; setting  each scalar target value to either 1 or 0, indicating which class the corresponding input belongs to (0- no risk, 1 –risk in action). For instance, it is possible to define the exclusive-or classification problem as follows:

inputs = [0 1 0 1; 0 0 1 1];

targets = [0 1 0 1];

Alternately, target vectors can have N elements, where for each target vector, one element is 1 and the others are 0. This defines a problem where inputs are to be classified into N different classes.

Classification problems involving only two classes can be represented using either format. The targets can consist of either scalar 1/0 elements or two-element vectors, with one element being 1 and the other element being 0.

 

 

3. OCCUPATIONAL RISK PREDICTION AND PREDICTORS

PRAT is based on a combination of expert systems with Bayesian approach and with tools to develop efficient training on the basis of risk analysis, as it could be seen in the figure...

.

Figure 1 PRAT metodological structure

The essential points of the risk analysis conducted according to the predictive Bayesian approach are identification of observable risk values, prediction and uncertainty assessments of these quantities, using all the relevant information. The risk analysis summarizes the knowledge and lack of knowledge concerning critical operations and other activities, and give in this way a basis for making rational decisions.

 

 

3.1.PREDICTIVE RISK ANALYSIS

 

Predictive Risk Analysis (PRA) is an efficient and optimal way to identify and evaluate risks that could action at the workplace, even if these risks have not yet an imprint on employees like occupational incidents and accidents.

PRA uses two kind of predictors:

-linear predictors- that could be classified mainly in two categories:

            -human reliability and failures;

            -system reliability and failures;

-knowledge nucleic predictors- that are based on structural knowledge about specific activity phases, arranged on root cause- possible effects structure;

3.1.1. Linear predictors

As a prerequisite for PRA ,a detailed data collection is required in order to have the materials for the building of linear predictors Linear predictors are extracted from the enterprise data. If the enterprise has a well established system of control and supervision then all the data could be found there. If not, some data must be specially colected for the linear predictors.

.Linear predictors should be used having a minimum of 6 months data. Such predictors can be:

-number of unexpected events/shift;

-number of unexpected events/month/year;

-identified root causes/shift/month/year;

-average cost of an unexpected event;

-number of minor injuries/month/year;

-average loss/month/year;

-operator load/shift;

-operator behaviour/performance/shift; some of these predictors could be measured or counted; others, like the performance of the operator during the shift could be evaluated by the supervisor using a simple Likert scale where 0 means non performance and 5 means excellent performance.

 

 

3.1.2. Knowledge nucleic predictors

Such predictors are built around knowledge seeds, taking into account a root cause- effect structure.

Among such predictors there could be defined:

- Work demands have generally been defined as referring to a set of prescribed tasks that an individual performs while occupying a position in an organization; if a task could be completely described by best practice procedures the work demands predictor should tell what could go wrong in the specific activity, taking into account the most frequent root causes. Research suggests that when job demands require “too much” effort and time (i.e. deadlines are too tight,resources are insufficient to allow the employee to fulfill responsibilities at work during regular hours), energy and time resources are depleted. Over time, high job demands

have been found to build up and hamper one’s ability to function outside of work (i.e. to fulfill one’s obligations to spouse, children, elder parents, community)

-Occupational (safety) culture: Workplace culture refers to a deep level of shared beliefs

and assumptions, many of which operate below the conscious level of those who are members of the culture (Lewis & Dyer, 2002). Excessive work demands are rarely a formal part of the employment contract (Lewis & Cooper, 1999). Rather, they often reflect the informal job expectations that are part of the organizational culture.There could be identified three operational levels of organizational culture: artifacts, values and assumptions.

Formal policies can be considered as artifacts—the surface level indicators of an organization’s intentions. Unfortunately, these formal intentions may be blocked by

“counterproductive” values and assumptions (Lewis & Dyer, 2002) and the supportive policies in place within the organization remain unused.There is a link link between having a “family-friendly” culture and being able to balance competing work and family demands. The

authors (Guerts  2003) distinguished between having family-friendly policies (i.e. formal arrangements that are provided) and actually being “family friendly” (i.e. the supportive attitude of supervisors and colleagues toward the use of these arrangements). The missing link, they contended, is the organizational culture. They noted that for “family-friendly” policies to have their desired impact with respect to promoting work–life balance, the use of these policies must be respected and accepted within the organization. In other words, the unwritten rules and norms of the organization must support balance for the policies to succeed.

-Behavioral patterns: the key element of the prediction is how would the employee behave in an unexpected situation. This could be very different from  country to country and occupational culture to occupational culture. However, there could be built behavioral pattern matching models that could give a general ideea regarding the would be behavior of a person in a specific context. A necessary feature of effective risk assessment is the identification of variables contributing to and sustaining an individual’s involvement in risk based behavior. Much research has been dedicated to this task and has highlighted a number of important historical and psychosocial factors as relevant to the prediction of dangerousness and persistent risk behavior. Some major predictors are shown below:

-daily stress;

- antisocial attitudes;

- antisocial peer associates;

- interpersonal conflict,

-unstable work arrangements;

-demographic variables (being male, single, and younger in age);

- employment problems

Accurate assessment of risk is an essential step in the successful reduction of risk. The guiding principles underlying efficient risk assessment/rehabilitation are the risk, need, and responsivity principles

-The risk principle is based on the premise that risk prone behaviour can be predicted and that the intensity of intervention to reduce this risk should be matched to the offender’s risk level.

-The need principle recognizes that certain risk factors are capable of being changed in a manner that reduces risk. These “behavioral needs” relate to lifestyle, cognitions, and behaviour (e.g., antisocial attitudes, substance abuse).

-The responsivity principle is concerned with the style and method of intervention used to target safety needs. Essentially, the choice of treatment should be based upon empirically-supported programs for the reduction risk, such as cognitive-behavioural and social learning approaches. The intervention also should be sensitive to the worker learning style and other factors that may interfere with his or her ability to respond to the intervention, such as mental disorder, motivation to change, or physical impairments. Adherence to the risk-need-responsivity principles has been shown to contribute to greater risk reduction than interventions ignoring or minimally adopting these principles.  The extensive research on the prediction of risk and the risk-need-responsivity principles has provided meaningful guideposts from which to construct valid instruments for the purposes of violence risk assessment.

 

 

Results of PRA should show the most probable and severe risks that could happen at the workplace together with the scenarios that could occur taking into account specific workplace contexts. This would be the basis for decision processes taken by the management.

For example, in an enterprise   there were identified 400 root causes/year (corresponding to the same number of unpredicted events) regarding defective plug-ins into an office building. The most frequent incident concerned a plug-in that was not well fixed in the wall- so during an inadequate handling it was moved or taken out of the wall .PRA had shown that the most probable- and severe- risk was the risk of fire started by an electric failure- fire that, taking into account the work context- would spread quickly in the building. A detailed worst case scenario was generated in order to be as a case study for the predictive training. The correspondent predictive training was focused on the following main aspects:

-check-up of all the plug-ins with qualified electricians and fix them in their encasement ;

-training the employees to work carefully with the plug-ins, especially with the unstable ones;

-elimination of fire sources (like paper) from the desks and from the office;

-stopping the risk (fire) at the source upon an immediate action with appropriate means;

-evacuating quickly and in order of the personnel;