Car Fuel Consumption Part I - Analysis before Predictions 🚗📈🔮

This is the first part of a project series to show a case in which I will apply advanced analytics, involving data visualization and predictions with Machine Learning, regarding Car Fuel Consumption.

1. Project description 👇

Part I: Exploratory Data Analysis (EDA) Visualization with Python.

For this project, I used a "Car Fuel Consumption per gas fuel type" dataset from Kaggle. From the Kaggle Dataset context, there are 4 questions formulated:

• Question #1: Which gas type consumes the most? E10 or SP98?
• Question #2: How much is the consume?
• Question #3: It consumes 0.4 liters more with E10 gas, isn't it?
• Question #4: Which of the two fuels is cheaper, E10 or SP 98?

"All these questions are answered in the notebook of this repository."

Exploratory Data Analysis => Summary:

• Performing data maintenance or cleaning: Duplicates, null-values, outliers
• String operations: Normalize to lowercase, replacing numbers to categorical and vice-versa
• Feature Engineering: Merging features, such as 'consumption rate' based on distance and consumed gas
• Relational model transformation: For future project steps
• Answering questions from our Exploratory Data Analysis: The main goal of this repository

Highlights:

Question #1: Which gas type consumes the most? E10 or SP98?

We can observe the consume rate (liter per kms), regarding the car speed records, is higher for the gas type SP98

Question #2: How much is the consumption?

Code Snippet:

Stats Overview

gas_type	count	mean	std	min	25%	50%	75%	max
E10	138.0	0.304058	0.125085	0.04	0.22	0.295	0.3975	0.68
SP98	197.0	0.321421	0.126652	0.09	0.24	0.310	0.4000	0.73

Mean Analysis

gas_type	consume_rate	speed
E10	0.304058	43.289855
SP98	0.321421	41.497462

Median Analysis

gas_type	consume_rate	speed
E10	0.295	42.0
SP98	0.310	41.0

• E10 has a mean consumption rate of 0.304 and a median of 0.295.
• SP98 has a mean consumption rate of 0.321 and a median of 0.310.

Both gas types show similar patterns in terms of distribution and spread (std), with SP98 having a slightly higher average consumption rate compared to E10.

Question #3: It consumes 0.4 liters more with E10 gas, isn't it?

Here is the reason why I created a relational-ish table df:

Code Snippet:

I am showing in the previous image, the merge of the gas consumption records to the treated dataset, in order to figure out if E10 is consumed 0.4 liters more than SP98. Now, with a groupby, I am getting the mean and the median for such consumptions:

Mean Analysis

gas_type	consume
E10	4.781159
SP98	4.668020

Code Snippet:

Median Analysis

gas_type	consume
E10	4.7
SP98	4.6

Code Snippet:

It consumes 0.4 liters more with E10 gas, isn't it? The answer is Not necessarily.

It depended on the number of experiments per gas type. That is why I did the Feature Engineering in the first place, to mean creating the column 'consume_rate' to be accurate in telling which gas type consumes the most.

Question #4: Which of the two fuels is cheaper, E10 or SP 98?

From the Kaggle Dataset context:

• E10 is sold for € 1.38
• SP98 is sold for € 1.46

Code Snippet:

Comparative Price Analysis between E10 and SP98

In this analysis, we compare the mean and median prices paid for two types of gasoline: E10 and SP98.

Mean Price

The mean price is the average of all prices paid for each type of gasoline. Here are the results:

• Mean price of E10: 6.598
• Mean price of SP98: 6.815

This means that, on average, the price paid for E10 is lower than the price paid for SP98.

Is E10 cheaper than SP98? Yes, on average, E10 is cheaper than SP98.

Median Price

The median price is the middle value in a set of data when it is ordered from lowest to highest. Here are the results:

• Median price of E10: 6.486
• Median price of SP98: 6.716

This means that if we order all the prices paid for each type of gasoline, the central value for E10 is lower than the central value for SP98.

Is E10 cheaper than SP98? Yes, according to the median prices, E10 is cheaper than SP98.

Conclusion

Both the mean and median price analyses indicate that E10 is generally cheaper than SP98. This information can be useful for consumers looking for more economical options when choosing the type of gasoline for their vehicles.

After this Exploratory Data analysis, the next experiment will be Machine Learning training models for Gas Consumption predictions. We export the 'pre-processed' dataset as follows:

2. Technology stack 💻

Programming language:

• Python

Python Libraries:

• pandas: For data manipulation and analysis.
• numpy: For mathematical operations and array manipulation.
• matplotlib.pyplot: For data visualization.
• seaborn: For statistical data visualization.
• re: For regular expression operations.
• scipy: For scientific and technical computing.

Distribution platform

• Anaconda

Computing environment

• Jupyter Notebooks

3. Folder structure 📁

└── project
    ├── data
    │   ├── raw
    │   │   └── measurements.csv
    │   └── pre_processed
    │   │   └── pre_processed_gas_df.csv
    ├── notebooks
    │   └── main.ipynb
    └── README.md    

4. Next steps 💡

Predictions

• Do you have any hypothesis?
• Can you make any kind of prediction: regression and/or classification?

Enriching the dataset

• Obtain related data by web scraping or with APIs.

Database

• Load the processed information into a database

Contact info📧

For further information, reach me at [email protected]