Restaurant Data Analysis Case Study: Selecting Competitive Categories and Optimal Locations with Python and QGIS

Project background – The exercise uses a restaurant dataset (96,398 records, 10 features) to practice data‑driven decision making with Python and QGIS.

Problem definition – (1) Identify the most competitive food category; (2) Compute a comprehensive score to select the best address for that category.

Data understanding – After loading the CSV, data.info() shows 10 columns of mixed types (float64, int64, object) and a small amount of missing values.

Data cleaning – data.isnull().values.sum() reveals 283 nulls (0.29%). Because the proportion is negligible, rows with nulls are dropped using data.dropna() , resulting in 96,255 clean records.

Feature selection – For the competitiveness analysis, five columns are kept: ['类别', '口味', '环境', '服务', '人均消费'] . Records with zero scores or zero consumption are filtered out.

Feature engineering – A new column 性价比 (price‑performance) is created by summing the three rating columns and dividing by 人均消费 (average spend), yielding a score per yuan.

Model building – grouping – The cleaned DataFrame df (54,886 rows) is grouped by 类别 and the mean of each metric is calculated.

Outlier handling – Box‑plot visualisation is used to detect outliers; an outlier‑removal function deletes extreme values before scoring.

Scoring formula – A weighted formula (e.g., 口味:人均消费:性价比 = 2:5:3) combines the three indicators into a final score. The three intermediate DataFrames are merged with pd.merge and the final score is computed.

Result – The category "甜品" (dessert) obtains the highest score, indicating it as the most competitive.

Location analysis – Using QGIS, layers for population density, road density, restaurant density, competitor density, longitude and latitude are prepared. Missing values are filled with zero ( dealdata.fillna(0, inplace=True) ) and the features are standardised. A composite index is calculated with weights 4:3:2:1 for the four density metrics.

Visualization – Matplotlib scatter plots and Bokeh interactive maps illustrate the spatial distribution of the composite index and highlight optimal sites: (121°472′E,31°301′N), (121°473′E,31°274′N), (121°493′E,31°244′N).

Conclusion – Opening a dessert shop at the identified coordinates maximises the combined score of market demand and competitive advantage.