Key Performance Indicator Analysis By MIC And LASSO And Rule Ensemble Prediction In Tianjin Mobile

Performance management, developed from performance appraisal with the pur-pose of the year’s end distribution, is hoping to comprehensively improve enterprisemanagement level. Since2002key performance indicators (KPI) has been locatedin the core position. The executive decision-making doesn’t work well without KPI.Monthly operational analysis is built around the KPI. Taking Tianjin Mobile Companyas an example, a system of performance appraisal is almost constructed, but the stafscore is generally low. Low scores, owing to high KPI targets, cannot give full playto its incentive efect. All kinds of correlation coefcients and LASSO are originallyapplied to formulate reasonable KPI target.Thus achieve the incentive to the customermanager and improve the economic benefits of mobile company.Customer manager’s performance is afected by three factors: external industry en-vironment, their own ability and random factors. The external business environment isrepresented by the company’s operating income. Correlation coefcients are developedto test their dependence. Linear correlation coefcient has more power under normaldistribution. Otherwise DCC(Distance Correlation Coefcient), HHG distance(HellerHeller Gorfine Distance) and MIC(maximal information Coefcient) are set up to for-mulate dependence between the external business environment and customer managerperformance.Employees age, gender, job title, job rank, the political landscape, the highestrecord of formal schooling and working days seven indicators are selected as indepen-dent variables, the total group informatization income of customer manager in2013asthe dependent variable. The most influencing independent ones are find out by LASSO.Gender, job title, job rank, the political landscape, the highest degree are classifica-tion variables and need virtual code before LASSO regression. LASSO loop iterativesteps can be optimized by Cross Validation. individual ability most afecting customermanager performance comes out.According to above analysis result, the customer manager performance has highrelevance on the company’s operating revenue. Company income need to be forecastbefore setting performance target. Mobile company customer erosion leads to incomeinduction for stif competition from China unicom and telecom.200group users in Tianjin binhai branch are taken as example. Signing time,2012whether in the network,2012customers,2012members of the group unified paid communication income,2012value-added business income are viewed as independent variables,2013whether inthe network as the dependent variable. Classify fitting is finished by Ensemble Rule,a popular algorithm in machine learning. Variable importance, partial correlation andvariable interaction are figured out. Group users tending to lose are predicted and needto pay key research so as to ensure the stable income growth of mobile business.