**5. Conclusion**

This paper investigated the impact of various leadership styles (transactional, transformational and laissez faire leadership styles) on leadership outcome. The leadership outcome was measured using extra effort, effectiveness and satisfaction. A number of statistical and data mining methods were used. The statistical methods included were correlation and regression analysis, whereas rule-based and decision tree algorithms were part of data mining algorithms used to extract interesting information from the data. Multiple regression analysis of independent variables (leadership styles) and dependent variables (extra effort, effectiveness and satisfaction) was carried out. The results indicated that transformation leadership style had a significant relationship with all three dependent variables. Moreover, the regression analysis of leadership outcome (aggregation of extra effort, satisfaction and effectiveness) and three leadership styles were carried out, and experimental results reiterated that transformational leadership style had a significant relationship with leadership outcome.

The data mining algorithms (rule-based and decision tree algorithms) were used to extract actionable information from the surveyed data. The values of dependent variables (extra effort, effectiveness and satisfaction) were transformed from numeric to nominal (low, medium and high). The single feature-based rules generated by OneR algorithm suggested that transformational leadership style is the most important variable to split dependent class variables. The final obtained models had accuracy of 66, 63 and 60% on extra effort, effectiveness and satisfaction, respectively. The rules extracted from this algorithm can be used to devise various management policies to enhance supervisors' leadership outcomes (extra effort, effectiveness and satisfaction). Another algorithm used in this thesis to analyse surveyed data was Modlem. This algorithm produced total 72 rules with 91% accuracy. These rules can be used to assist management to make rationale and more informed decisions. Lastly, J48, a decision tree-based algorithm, was used to construct decision tree. The main contribution of this research is that we have used data mining techniques to extract hidden and interesting patterns in the data in the form of understandable rules and decision trees.

Input, process, output and feedback are four main components of a standard business information system (**Figure 4**). In the context of this research, inputs were the data used (including independent and dependent variables) and processes were statistical and data mining

**Figure 4.** Components of the business information systems.

algorithms applied on the data (correlation, regression, OneR, J48 and Modlem). The outputs were correlation, regression and decision trees obtained using data mining algorithms. The feedback consisted of the information extracted from the regression analysis and various rules extracted from the data mining algorithms. Management can use direct and actionable output from various processes to direct their various internal stakeholders. In the light of this feedback, internal stakeholders can improve their productivity and performance.
