The World Bank estimates that the global poor are living in a rural environment, young, lacking education and with a large family with many children [3] . Economic Poverty is defined as a person living under 1.90 USD per day or suffers an economic deprivation [4] . Sociologists will associate poverty with health, education or safety deprivation. All the member states of the United Nations (UN) have people living in poverty. Some states offer an environment where poverty is more present than other states. The United Nations classifies countries into three categories: Developed economies, economies in transition and developing countries [5] . Several reports will define poverty as the Gross National Income per capita levels for each of those three categories. We intent to use a methodology that is based on statistical data and an operational research methodology named Dominance-based Rough Set to determine this classification and propose to define poverty from four different perspectives: political poverty, economical poverty, sociological poverty and technological poverty.

A bunch of artificial intelligence tools and algorithms exist in order to help decision makers and leaders defining strategy and courses of action. One of them is the Rough set theory, which was developed by Pawlak [6] [7] and by Pawlak and Slowinski [8] . It is a mathematical tool that is used to support decision-making processes in fields such as medecine, banking, engineering, learning, location selection, pharmacology, finance, market analysis and economics [9] - [17] . It was later modified by Greco, Matarazo and Slowinski [10] and renamed the “Dominance-based Rough Set Approach” (DRSA). And then, Zaras developed it for mixed data (deterministic, probabilistic and fuzzy) [11] . Because of its advantage of sorting decision rules explaining the classification of objects, this research uses DRSA. The purpose is one of developing sound strategic objectives for all Third world countries in order to help decision makers, leaders, non-governmental organizations, help funds and other international organization or interest groups target specific objectives to improve the economical, political, sociological and technological situation of the Third world. A total of 23 indicators were selected by experts and categorized in four different perspectives (Political, Economical, Sociological and Technological).

Section 2 presents the statistical analysis which is a list of all the significant correlations between all the indicators classified per perspectives (PEST). Section 3 shows the application of DRSA to classify all the countries member of the United Nations, with regard to the perspectives (PEST) and explains the decision rules for each category (countries classified as: A, B and C). Section 4 proposes a strategy map for each individual country, proposing strategic objectives and performance measures to improve and monitor the sustainable development of all the selected Third world countries.

This section develops the application of the Dominance-based Rough Set Approach (DRSA) in order to determine the strategic objectives of each United Nations countries and improve their overall classification. First step is to classify all the Countries per perspectives in category A, B or C: Category [A] Countries that are doing well according to the selected indicators; [B] Countries that need support to acquire category A status; [C] Countries ranked the lowest and meeting special support with regard to the criterion or criteria considered. Appendix B presents the overall evaluation of all the United Nations countries with respect to the four conditional criteria as determined on the basis of each perspective

(PEST) and with respect to the decisional criterion. Table 6 is a summary of all the classifications per Continent and Sub-Regions. Second step is to extract decision rules for all the variables on a first time, and on a second time individually on each perspective (PEST). Third step, for each country, we could determine and prioritize its strategic objectives with regard to their respective variables and values.

Table 6 demonstrates each continent separately, their sub-regions (if required) and the number of countries per classification. Finally, the last column explains the number of people per sub-region and the percentage of the population compared to the world’s population.

The ranking of the 193 countries members of the United Nations was perform on the basis of the 23 criteria measured by 23 indicators. The same was also done for each perspective on the basis of each respective criteria. That kind of problem can be represented using the AXE model, where:

A is a finite set of countries a_i for i = 1 , 2 ⋯ 193 ;

X is a finite set of criteria X_k for k = 1 , 2 , ⋯ , 23 or X_kj for k j = 1 , 2 , ⋯ , n j for each perspective j.

E is the set of evaluations measured by indicators e_ik with respect to criterion X_k or indicators e_ikj with respect to criterion X_kj for each perspective j.

Since the weights of indicators are assumed equal, we performed the weighted average rank method in order to rank countries. In fact, the countries are ranked from the most to the least preferable in terms of each indicator in relation to each criterion. Calculation for each country of its weighted average rank were performed to obtain the rating of the countries with respect to a given perspective but also on overall classification of the 23 criteria.

For each perspective j, the weighted average of country i,

r i j = ∑ k j w k j r k i j (1)

The overall weighted average of country i,

r i = ∑ k w k r k i (2)

where:

w_k is the weight of criterion k and w_kj for perspective j;

r_ki is a rank of country i with respect to criterion k and r_kij for perspective j.

With the obtained ranking of 193 countries, overall and for each perspective, countries were classified into three categories A, B and C (Appendix B).

In rough set theory, the decision problem is represented as a table, the rows corresponding to objects and the columns to attributes (see Table 7). In our approach, the objects are the countries and we used two types of attributes: conditional and decisional. In according to ranking obtained by using multi-criteria method we classified countries to three categories A, B and C. The decision in the decision table with respect to the decisional attribute takes one of three values: Country belongs to the category A, B or C.

The remaining attributes will be called conditionals and these will be from our multi-criteria AXE problem (the n_j criteria, in the case of overall classification 23 criteria). With respect to each conditional attribute, the evaluation of the country takes the value of the indicator e_kij in relation to each criterion k.

Based on the approximations obtained by means of the dominance relations, it is possible using DRSA methodology to derive a generalized description of the preferential information contained in the decision table, in terms of decision rules.

Decision rules were extracted using the 4eMka2 software [22] , developed by the intelligent decision support systems laboratory (IDSS) at the computing science institute of the Poznan University of Technology. Table 8 presents the combination of decision rules for all the perspectives combined and Table 9 presents

each combination of decision rules for each perspective individually. It is important to mention that all the rules that are presented have a minimum relative strength of 20% and limited to 4 conditional criteria.

Decision rules 5, 6 and 7 help us determine poverty when all indicators and perspectives are combined. Therefore, for all the states within the United Nations, these are the following decision rules and conditions for poverty:

1) The Gross National Product is equal or lower than 917.56$ per capita;

2) The number of secure internet servers is equal or smaller than 5.2 per million users and the conflict index, which represents the number of battle deaths from internal conflict between at least one government armed forces is equal or greater than 2.03; and

3) The corruption perception index is equal or lower than 28 with a GNP equal or lower than 1642.73$ per capita.

Decision rules 9 and 10 dictate what the conditions for political poverty are:

1) Conflict index, is equal or larger than 1.26 combined with military expenditures in % of GDP equal or larger than 2.18%; and

2) Conflict index same as the first example (1.26) combined with the global competitiveness index equal or lower than 4.35.

Decision rule 18 dictates that economical poverty can be defined as these following sets of conditions:

1) GNP per capita is equal or lower than 738.64$ (USD) per capita combined with exports of merchandise per capita equal or lower than 200.74$.

Decision rules 23 and 24 give us the sets of conditions for sociological poverty:

1) Life expectancy for women equal or less than 68.79 years, urban population is equal or less than 59.79% of the entire population of the state and school life expectancy is equal or less than 13 years; and

2) Urban population is equal or less than 35.04%, school life expectancy is equal or less than 10 years and with an adolescent fertility equal or over 42.75 birth per 1000.

It is important to state that the sociological perspective has three core attributes that impact the most the classification of the states: School life expectancy, % of urban population and adolescent fertility index per 1000.

Decision rules 31, 32 and 33 define technological poverty with the following conditions:

1) The number of secure internet servers is equal or smaller than 3.2 per million users and the number of internet user is equal or smaller than 31.87% of the population;

2) Fixed broadband internet is equal or less than 2.79 per 100 people, production of academic papers is equal or less than 5.71 articles per million people, cellular subscriptions is equal or less than 124.94 per 100 people and secure internet servers is equal or less than 62.87 per million people; and

3) Cellular phones subscriptions is equal or less than 76.37 per 100 people, fixed broadband internet is equal or less than 0.54 per 100 people and secure internet servers is equal or less than 15.64 per million people.

Several of the indicators selected are results. For example, GNP per capita cannot be increased or decreased voluntarily and decision makers have no or very little power in influencing the results. Other indicators, such as school life expectancy, adolescent fertility, corruption perception index, military expenditures and the number of academic papers are indicators where decision makers may influence by changing policies, targeting funding or creating programs to educate the population with regards to these factors. As proved with the correlation matrix, since economical and technological indicators are correlated with sociological and political indicators, it is plausible to set objectives with specific targets for each country categorized as C and dictate what needs to be done to improve their situation. All the decision rules equal or greater than A and decision rules equal or greater than B may be either used for prediction or transformed into strategic objectives with specific targets for each country. As an example, we selected randomly a country classified as C, the state of Cambodia. Again, any country in category C should have interest to reach the threshold pointed out in decision rules in order to reach category B or even A. However, their improvement objectives would be different since their actual performance is different. Table 10 describes that decision rules equal or greater than B are transformed into strategic objectives and targets for the randomly selected Cambodia. You will find all the strategic objectives and targets for the decision rule 22 for all the countries classified as C in Appendix C.