You are on page 1of 4

DigiTech Portal Assessment

FY23

Technical
Requirement
Specification

Project Id
Project Name Sales Forecast Application
Organization Kaar Technologies
Manager, Email Girish M.P, mpgirish@kaartech.com
Document Author, E mail Hariesh G, ghariesh@kaartech.com
Hemchandhar R, rhemchandhar@kaartech.com
Tharachandar B, btharachandar@kaartech.com
Mithran B, bmithran@kaartech.com
Nishanthan V, vnishanthan@kaartech.com
Logesh Karthik S, slogeshkarthik@kaartech.com
OBJECTIVE 1

Create a Web Application to perform Sales Forecasting with given


dataset. Application should have features to extract data, cleanse data and
do prediction. The prediction results should be shown in interactive
dashboard.

Tech Stack
SPECIFICATIONS 2

Building Pipeline:
Build a flexible and fully automated Machine Learning Pipeline. The pipeline
should perform task from Data Extraction to Prediction.

Steps:
• Data Extraction
• Data Cleaning
• Pre-processing
• Training Model
• Prediction
• Visualisation

Data Extraction:
1. User should upload sales dataset as CSV file through Angular App.
2. User will have option to set target, features, and time period of future
forecast.
3. Using Flask API Endpoint, the uploaded dataset will be passed on to backend.

Data Cleaning:
1. Before Prediction data needs to be cleaned to gain better result.
2. Fixing or removing incorrect, corrupted, incorrectly formatted, duplicate, or
incomplete data can be done in this step.
SPECIFICATIONS 3

Pre-processing:
1. Raw data should be transformed according to the requirement.
2. Data validation should be carried out. Process like Normalisation,
Standardisation, and other EDA (Exploratory Data Analysis) steps should be
done in this step.

Prediction:
1. Predict the target feature for time period selected by User.
2. Create Actual vs Prediction comparison to display in the Application.
3. Find model validation metrics like RMSE, MAPE, Accuracy.
4. Using Flask API Endpoint pass the data back to Angular App.

Visualization:
1. Display the results of prediction in the form of graphs in Angular App.
2. Also Build an informative dashboard based on the prediction result in Power
BI.
a. Store the Actual vs Prediction table as CSV file and use that as
datasource for Power BI Dashboard.
b. The Power BI Dashboard should be more Business oriented.

You might also like