Team 68_Forecasting Model Singtel

Forecasting Model Singtel
Production Gallery

1. Inspiration

Singtel has always prided itself on being a company that focuses on its customer's needs and makes sure to deliver mobile devices on time to our customers when they request them. It is essential to develop a forecasting model that considers the changes in demand (such as promotional campaigns or festive seasons) and order lead times required by suppliers to forecast future mobile device demand. Hence, we embarked on a project that aims at creating a model that can be used to predict future sales based on past trends and current market conditions, along with any upcoming or ongoing promotional campaigns that may be coming up.

2. What it does

Our model can predict the future demand for the products in the next ten weeks based on factors that can affect the market, such as promotional campaigns, new devices released, and festive seasons. Ultimately, we provide a dashboard to give stakeholders a clear vision of the forecasting results and trends.

3. How we built it

At first, we converted the raw data into a format arranging the readable and viable data. We continued to train many models (Simple Exponential Smoothing (SES), Holt Winter’s Exponential Smoothing (HWES), AR, ARIMA, Auto ARIMA, SARIMAX, Bayesian regression, Lasso Regression, Randomforest, Lightgbm, XGBoost, SVM RBF, KNN, LSTM, Grid search - SVM, Bayesian processes - Xgboost, Polynomial Regression) with different results and finally chose the most suitable model (Polynomial Regression) for this dataset. In the end, we created a dashboard on our website.

4. Challenges we ran into

This project initially caused us many problems since we were unfamiliar with the data science field. This meant that we had to do a large amount of research on a variety of models in order to identify the best one for the dataset. We also had to spend a lot of time preprocessing and cleaning the raw data since it was too raw to be analyzed efficiently.

5. What we learned

Data is the future. The benefits of big data cover both the internal operation process and the optimization of the customer experience journey. More than diamonds with a fixed value, big data is an invaluable resource for businesses that know how to mine correctly.

6. What's next for Singtel

6.1. Build "sure to win" business plans

Using data, many plans to enter the market, scaling or introducing new products/services no longer come with many risks. Big data has estimated and polled the market in advance. As a result, businesses can perfect their products/services and avoid all previous dangers. In addition, data helps companies get the right strategy, especially in determining a price frame that is in balance with the market. Since people generally like to maintain habits, any change will receive many reactions. And big data is the springboard to bring these plans closer to a positive response, extending the winnings.

6.2. Understanding customer needs, anticipating changes

Big data helps businesses name customer needs before they realize them. As society develops, people also change over time. Companies can accurately identify new needs arising in this development process through big data. As a result, businesses soon grasp tastes, prepare new products/services, and quickly respond to customers.