The Fifth Elephant 2018

The seventh edition of India's best data conference

A Time Series Analysis of District-wise Government Spending

Submitted by Gaurav Godhwani (@gggodhwani) on Saturday, 31 March 2018

videocam_off

Technical level

Beginner

Section

Full talk

Status

Submitted

Vote on this proposal

Login to vote

Total votes:  +7

Abstract

About District Treasuries: District Treasuries are the nodal offices for all financial transactions of the Government within the district, managing both payment and receipts. They also monitors the activities of various sub-treasuries which work as an extension of the Treasuries at the Tehsil/Taluka level. Each district has various Drawing & Disbursing Officers who are authorised to draw money can present their claims in the Treasury which are then accounted by concerned authorities. Various states in India have developed Integrated Financial Management System which publishes detailed information of daily transactions happening at district treasuries within a state.

About Time Series Analysis & Inferences: The detailed information of daily transactions at district treasury can helps us perform near real-time tracking of flow and utilisation of funds. This can be used to track expenditure on various schemes and social sectors, anomalies in fund disbursement, understanding near real-time alerts and predicting timely utilization of budgets. In this talk, we will explore how we can harness time-series modelling and analysis to better understand functioning of various district treasuries in India.

Outline

The session will be organized as:

  • Setting the scene
  • Getting the District Treasuries Data
  • Exploratory Data Analysis of the District Fund Flow
  • Evaluation of various Time Series Algorithms
  • Visualizing trends and the big picture
  • Scaling Up
  • How you can contribute
  • Future
  • Questions

Speaker bio

Gaurav Godhwani is working as the technical lead with Centre for Budget and Governance Accountability, majorly focussing on Open Budgets India initiative. He has served as co-founder and chapter leader for DataKind Bangalore, building a team of pro-bono data scientists to help nonprofits start projects addressing critical humanitarian problems and explore how data science can be applied to solve these challenges. He has previously led data projects at Goibibo and PromptCloud, later focusing his work towards data-for-good ecosystem. He is a passionate open data and open source contributor.

Links

Slides

https://docs.google.com/presentation/d/17Suj2GZhgkcgpdVWeuMOBGFXXpmCn5vprEk-71sfx70/edit?usp=sharing

Comments

  • 1
    Zainab Bawa (@zainabbawa) Reviewer 8 months ago

    Thanks for this proposal Gaurav. What other options did you evaluate for solving your problem? Why was time-series analysis best suited for this use-case?

  • 1
    Gaurav Godhwani (@gggodhwani) Proposer 7 months ago

    Thanks Zainab, since the data is of time-series nature, we preferred to use time-series algorithms over other statistical techniques to take into account factors like seasonality, etc. More about this is covered in my draft slides and always happy to discuss in detail. Would be working to add more data and experiment results in coming days.

Login with Twitter or Google to leave a comment