Intelligent Cash Flow Forecasting

Inspiration

Cash flow forecasting (CFF) allows businesses to estimate their cash positions in future to ensure better liquidity management and take efficient borrowing and investment decisions. The accuracy of the CFF is very critical as this drives the key business decisions for the startups and large companies.

Corporates are facing numerous challenges such as Inconsistent and Bad Data, Manual adjustments negatively impacts the actual narrative provided by historical data, Lack of standard templates & processes, Time consuming and tedious data collection and consolidation activities, etc.

Because of these challenges they often find difficult to arrive at accurate forecast and make appropriate decisions. This motivated us to address this problem by developing a solution which provides real time and accurate forecast built using machine learning models.

What it does

Intelligent cashflow forecasting solution by HCL aims at leveraging AI/ML algorithms to understand the historical data and arrive at accurate forecasts. This solution can be implemented as an extension to Fusion Cash Management which is currently implemented by banks to offer wide range of cash management services to Businesses. Once integrated with FCM, Businesses will have the access to AI driven cashflow forecasting, a hassle-free way of understanding their future cash positions.

How We built it

Three key pillars of the solution is Data Analysis & Pre Processing, Model Selection, Train & Score the Model to derive the right output Data Ingestion identifies the right data set and parameters required for the model which drives the selection of the model and the accuracy of the output. Exploratory Data Analysis is performed to analyze and understand the data better to employ the right models. For this forecasting the Prophet, XG Boost, LSTM and Timeseries models are selected. In this solution all the models are applied to select the best champion model based on the accuracy of the forecast. Training and testing the model on the various data sets prevents over fitting of the data and shows how well the model can learn and adapt to the new sets of data. Once the models are trained and tested, running it on the data sets provides the forecasted output for the Payables, Receivables and cash flow.

Challenges We ran into

Obtaining test data to train and test the model was a major challenge. The difficulty was overcome by fabricating the data which was done keeping in mind various aspects of the business domain.

Accomplishments that I'm proud of

Identifying the right parameters and developing an intelligent model using champion selection paradigm.

What I learned

Getting the right data is key to develop a model. To arrive at accurate forecasts, we must identify the right set parameters based on the business domain.

What's next for Intelligent Cashflow Forecasting

This solution does not stop only with the cash flow forecasting and opens up lot of potential opportunities for Businesses. Simulation & What if analysis can be performed based on the forecast output. Identify the unobserved / unearthed patterns using the Payables and Receivables forecast. Understand the data better to improve the data quality and forecast accuracy.

Built With

• data
• machine-learning
• powerbi
• python
• science