This iPython notebook documents our Harvard CSCI E-109 2014 data science class project to explore, analyze, build model, and predict the demand for telecom repair jobs based on weather and time-related phenomena.
•Dan Traviglia, CSCI E-109 •Kaiss Alahmady, CSCI E-109 •Prabhu Akula, CSCI E-109 •Sundaresan Manoharan, CSCI E-109
Please see our Telecom-Job-Psychic for more information.
You'll need to install iPython and iPython notebook to use our project file. The best way to do this though is to just install Anaconda.
Telecom services, such as voice, data and video services are delivered on various telecommunication networks to subscriber's homes and businesses. Failures and operating problems associated with the network and network equipment cause service problems in which some of them can be fixed remotely and others require dispatch of technicians to the site to repair and fix the problem(s). Managing the available resources and their capacity in executing the repair jobs is quite challenging because the daily volume of dispatch needs (jobs demand) is largely unpredictable due to several factors such as the characteristics of the network and network equipment, nature of repairs needed, regulations, weather, seasonality, and characteristics of the labor force (such as union membership, skills, culture); among others. On the other hand, if extra resources and capacity is secured, the cost to secure these resources in addition to unplanned overtime cost can be substantial. If a lack of resources and capacity exists, then broken promises and negative customer experiences occur which typically affects customer satisfaction. The ability to predict the jobs demand in advance for immediate and near short terms can provide substantial benefits in optimizing capacity, reducing network downtime and improving customer satisfaction.
We sought to develop a statistical model to predict the daily and 21-day period demands for telecom repair services to help in optimizing repairs resources capacity, improve operational effectiveness, reduce overtime cost and improve customer services.
Our inital hypothesis is that the number of repair jobs is related to weather conditions plus seasonality or some other time related phenomena. This project will help us identify and develop a model for repair jobs based based on this idea.
We'd love to hear your thoughts on our project work! We've put lots of effort into this project and have completed our own exploration, analysis, and modeling of the data but there are more areas to explore and plenty of ideas to test.
Please contact us if you'd like to merge your changes into our notebook.