Predictive models for rental car prices | Data Analysis case study

  • Demand –Supply optimization
  • Cost Savings
  • Comparison Shopping
  • Good Communication and Customer Support
  • Competitive Advantage
  • Fraud Detection
  • Anomaly detection

Relying on a dataset of car hire pricing

Information from Australia we built predictive models that can forecast prices and the number of available cars at a given point in time.

The context

Our client decided to identify data-based insights regarding customers' predictive behavior for the car rental industry in response to heightened interest expressed by companies in data analytics and the growing reliance on data-driven decisions.

The opportunity

Through data analysis (predictive models), the client identified the opportunity to provide its clients with patterns, trends, and correlations for optimizing their processes, improving efficiency, and driving innovation.

The objective

With a dataset of car hire pricing information, we defined a specific data analysis objective for our client: to build predictive models that can forecast prices & the number of available cars at a given time.

4 predictive models | Technical highlights

The ability to predict car rental prices in Australia can be quite powerful for both insurance and car rental companies.

That’s why we employed 4 time series models in order to predict what the future prices will be.

We trained the models on a subset of our data and tested them against known prices to determine based on statistical measures what is the most accurate model:

  • The 4 models made price predictions based on the training set.
  • The predicted prices can be compared with the test prices and compute the difference in percentage between the real price and the predicted price for every day.
  • To determine the overall performance of the model we used the Root Mean Square Error and compute it for every model.
  • Good Communication and Customer Support.
  • We observed that the best model is SARIMA because it has the smallest RMSE, while Linear Regression is the worst model with the highest RMSA.
  • Onced we determined that our best model is SARIMA, we made a real future prediction for the next 7 days outside the dataset.

Example of how we select a machine learning model

Demand-Supply Optimization

Cost Savings

Comparison Shopping


Anomaly Detection

Demand-Supply Optimization

By accurately predicting rental car prices (through these predictive models), insurance companies can anticipate periods of high demand and secure rental cars in advance. This proactive approach ensures that they have an adequate supply of vehicles available when claimants require them. It helps avoid delays in the claims process and enhances customer satisfaction.

Cost Savings

Predicting rental car prices allows insurance companies to assess the potential costs associated with car rentals. By understanding the expected prices for rental cars, they can better estimate the financial impact of claims and allocate appropriate reserves. This enhances their risk management capabilities and helps them plan and budget accordingly.

Comparison Shopping

With the ability to predict car rental prices, consumers can compare prices across various rental companies and make informed choices. They can identify periods when prices are expected to be lower and select the most affordable option without compromising on their requirements. Predictions provide valuable insights into the market, enabling consumers to find the best.


Good Communication & Customer Support
With rental car price forecasts, insurance companies can inform claimants about the expected reimbursement amounts based on the forecasted rental car prices, ensuring transparency and managing customer expectations.

Competitive advantage
Having the ability to predict rental car prices gives insurance companies a unique competitive advantage. It sets them apart from competitors by enabling them to offer more accurate and fair claim settlements.

Fraud Detection
Insurance fraud is a significant concern for insurance companies. A price prediction algorithm can help identify suspicious rental patterns or abnormal pricing behaviors that may indicate fraudulent activities.

Anomaly Detection

We determined that some values are abnormal based on Time Series Models for Anomaly detection. Below, we can see that the spikes from the original series have been highlighted in orange, raising this way some anomalies each time the number of cars increased significantly compared to the previous values.

Data Analytics services we provide

Roweb Data Analytics Services

Data collection

We collect the relevant data for analysis.

Roweb Data Analytics Services

Data exploration

We filter & analyse data to identify your business needs.

Roweb Data Analytics Services

Data visualization

Transform raw data into visually compelling insights.

Roweb Data Analytics Services

Data modeling

We create AI and machine learning models to process your data.

Roweb Data Analytics Services

Automated frameworks

Creation of automatic frameworks capable to use models in real time.

Roweb Data Analytics Services

Data monitoring

Solutions to help monitor and analyze data: Grafana, Kibana, Prometheus.

Interested to get a custom Data Analysis for your company/products/industry?

Data Analysis for various verticals & markets

Recruiting & HR

Tourism & Hospitality


Finance & Insurance

Real Estate

Transport & Logistics