WEATHER PATTERNS ANALYSIS

AND PREDICTION

Student’s Name:D.Manish Kumar

Submission Date: 22/12/2024
 INTRODUCTION
Objective:
• Weather Patterns Analysis and Prediction from 01-01-2015
to 31-03-2017
• Analyze weather data to uncover trends and patterns.
• Predict rain presence using machine learning techniques.
• Provide actionable insights from data-driven analysis.

Dataset Overview:
• Number of Records: 1000+ (example value)
• Features: 7 key features (e.g., Temperature, Humidity, Visibility, etc.)
 TOOLS AND TECHNIQUES
Tools:
• Python (Pandas, Matplotlib, Seaborn, Scikit-learn)
• Jupyter Notebook

Techniques:
• Exploratory Data Analysis (EDA)
• K-NN Classification
• K-Means Clustering
 EDA: KEY FINDINGS
Key Findings:
• Temperature ranges from 12°C to 45°C.
• Median humidity is 77%, highest during rainy conditions (~82%).
• Visibility significantly drops (~1.5 km) during rain.
• Frequent Weather Conditions: 'Haze' and 'WNW' wind direction
dominate.
 EDA: KEY FINDINGS
The dataset includes several weather parameters such as humidity, temperature,
pressure, dew point and so on.Let's delve deeper into the weather captured in the data.
1. Weather condition:
• The most common weather condition observed is Hazy, occurring for about 492 days
followed by smoke.
• The least frequent weather condition was drizzle, rainy, scattered and clean which
appearing only one day.
• There are totally 15 different weather conditions were recorded by the dataset.

2. DEW POINT:
• The minimum dew point is 1 and the maximum dew point is 28
• The average dew point is about 16.64
• The most frequent dew point is 12 which was seen for 58 days.
 EDA: KEY FINDINGS
3. HUMIDITY:
i)The minimum and maximum humidity was 6 and 100.
ii) Average humidity: 36.34
iii) Most frequently seen humidity is 31.

4. PRESSURE:
i) The minimum and maximum pressure is 994 and 1026.
ii) ii) Average pressure:1007.74
iii) iii) Most frequent pressure: 1014.
5. TEMPERATURE:
i) The minimum and maximum temperature is 12 and45.
ii)The most common temperature is 35 which was over 70 days.
iii)Average temperature is 30.78
 EDA: KEY FINDINGS
6. VISIBILITY:
1) The minimum and maximum visible are 0.2 and 55.
2) Average visibility: 2.41
3) most frequent visibility:2
7. WIND DIRECTION:
i) The most frequent wind direction is WNW which came over 130
days.
ii) ii) The least occurred wind direction is SSW which was only for 4
days.
8. PRESENCE OF RAIN:
Rain is present only for 18 days out of 730 days.
 EDA: VISUALS -
TEMPERATURE TRENDS
• Graph: Line chart showing temperature variations over time.
 EDA: VISUALS - RAIN
PRESENCE VS VISIBILITY
• Graph: Box plot comparing visibility during rainy and non-rainy
conditions.
 EDA: VISUALS -
HUMIDITY DISTRIBUTION
• Graph: Histogram showing the spread of humidity values.
 EDA: VISUALS - WEATHER
CONDITION FREQUENCY
• Graph: Bar chart highlighting the frequency of weather conditions.
 METHODOLOGY - K-NN
CLASSIFICATION

Steps:
• Normalize dataset for uniformity.
• Calculate distances using the Euclidean metric.
• Identify the top K nearest neighbors.
• Predict class based on majority voting among neighbors.
 METHODOLOGY - K-
MEANS CLUSTERING

Steps:
• Initialize centroids randomly.
• Assign points to the nearest cluster.
• Recompute centroids as the mean of assigned points.
• Iterate until centroids stabilize.
 FLOW DIAGRAMS FOR
BOTH METHODOLOGIES
 RESULTS - K-NN
CLASSIFICATION
Results:
• Accurately predicted rain presence using features like humidity
and visibility.
• Evidence includes a confusion matrix and prediction results table.
 RESULTS - K-MEANS
CLUSTERING
Results:
• Data grouped into clusters based on weather conditions.
• Final centroids and scatter plot depict clustering outcomes.
 INSIGHTS AND
LEARNINGS
Insights:
• Rainy conditions correlate with higher humidity and lower
visibility.
• 'Haze' and 'WNW' dominate as frequent weather and wind
conditions.

Learnings:
• Visibility is a strong predictor of rain.
• Seasonal temperature variations were evident, ranging from 12°C
to 45°C.
 CHALLENGES
• Data Quality: Missing values required preprocessing.
• Computational Complexity: High memory usage for K-NN on
large datasets.
• Cluster Interpretability: K-Means results required domain
knowledge.
 RECOMMENDATIONS
• Improve dataset quality with automated validation during
collection.
• Use approximate methods like KD-Trees for scalable K-NN.
• Collaborate with domain experts for better clustering
interpretation.
 CONCLUSION
Recap:
• Analyzed weather data to identify patterns, predict rain presence,
and group similar conditions.
• Key insights include correlations between visibility, humidity,
and rain presence.
 BROADER IMPLICATIONS
• Findings support climate monitoring and disaster preparedness.
• Applications extend to aviation, agriculture, and logistics.
• Insights aid in climate monitoring, urban planning, and disaster
preparedness.
• Applications in sectors like aviation, agriculture, and logistics are
significant.
 REFERENCES
• Tools: Python (Pandas, Matplotlib, Seaborn, Scikit-learn), Jupyter
Notebook.
• Websites: Pandas Documentation, Scikit-learn Documentation,
Matplotlib Documentation.
• Books: 'Hands-On Machine Learning' by Géron, 'Python for Data
Analysis' by McKinney.