Session 4: Data Science: Machine Learning¶
Attending the Data Science Fitting class was not as I had expected, as the lessons involved many mathematical concepts and coding. I will try to keep track of my learning experiences with the help of my colleagues and online resources.¶
Attending help session from Mr. Jean Michel Molennar¶
I used Chatgpt to understand the machine learning functions in details. The following are the prompt used for learning.
Sigmoid¶
import numpy as np
def sigmoid(x): return 1 / (1 + np.exp(-x))
x = np.array([-2, -1, 0, 1, 2]) print(sigmoid(x))
Tanh Activation¶
Leaky ReLU¶
Future Predictions of Fire counts¶
In [2]:
import pandas as pd
# Load dataset
df = pd.read_csv("datasets/firecounts.csv")
# View first rows
print(df.head())
# Summary statistics
print(df.describe())
Year Fire Counts
0 2001 199
1 2002 127
2 2003 170
3 2004 219
4 2005 204
Year Fire Counts
count 24.000000 24.000000
mean 2012.500000 240.125000
std 7.071068 95.406823
min 2001.000000 97.000000
25% 2006.750000 169.750000
50% 2012.500000 226.000000
75% 2018.250000 282.750000
max 2024.000000 442.000000
Counts Predictions based on Year.¶
In [4]:
import numpy as np
from sklearn.model_selection import train_test_split
X = df[["Year"]] # Feature
y = df["Fire Counts"] # Target
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.2, random_state=42
)
Linear Regression¶
In [7]:
from sklearn.linear_model import LinearRegression
model = LinearRegression()
model.fit(X_train, y_train)
print("Model trained.")
print("Coefficient:", model.coef_)
print("Intercept:", model.intercept_)
Model trained. Coefficient: [-3.01486698] Intercept: 6300.923317683881
Forecast Future years (2025-2030)¶
In [8]:
future_years = pd.DataFrame({"Year": np.arange(2025, 2031)})
future_predictions = model.predict(future_years)
forecast_df = pd.DataFrame({
"Year": future_years["Year"],
"Predicted Fire Counts": future_predictions
})
print(forecast_df)
Year Predicted Fire Counts 0 2025 195.817684 1 2026 192.802817 2 2027 189.787950 3 2028 186.773083 4 2029 183.758216 5 2030 180.743349
Historical vs Forecast¶
In [9]:
plt.figure(figsize=(10,5))
# Plot historical data
plt.plot(df["Year"], df["Fire Counts"], marker='o', label="Historical Data")
# Plot forecast
plt.plot(forecast_df["Year"], forecast_df["Predicted Fire Counts"],
marker='x', linestyle="--", label="Forecast")
plt.xlabel("Year")
plt.ylabel("Fire Counts")
plt.title("Historical and Forecasted Fire Counts")
plt.legend()
plt.grid(True)
plt.show()
Image visualization¶
In [ ]:
I used Chatgpt promt to visualize the original image in different colors and sizes.
Original Image¶
In [18]:
from PIL import Image, ImageOps, ImageFilter
import numpy as np
import matplotlib.pyplot as plt
# Load image
img = Image.open("images/dzo.png").convert("RGBA")
# Convert to grayscale for masking
gray = ImageOps.grayscale(img)
# --- BETTER MASK ---
# 1. Lower threshold to catch faint lines (light blue)
# 2. Invert mask so lines become white (255)
mask = gray.point(lambda x: 255 if x < 230 else 0, mode="L")
# --- THICKEN LINES ---
# Apply dilation using a blur + threshold trick
mask = mask.filter(ImageFilter.GaussianBlur(1.2)) # smooth + thicken
mask = mask.point(lambda x: 255 if x > 30 else 0, 'L') # restore clean mask
# Function to recolor the CLEAN mask
def color_line(mask, color):
w, h = mask.size
out = Image.new("RGBA", (w, h), color=color)
out.putalpha(mask) # treat mask as alpha channel
return out
# Create colored lines
red_line = color_line(mask, (255, 0, 0, 255))
yellow_line = color_line(mask, (255, 255, 0, 255))
blue_line = color_line(mask, (0, 0, 255, 255))
# Display
plt.figure(figsize=(12, 4))
plt.subplot(1, 3, 1)
plt.imshow(red_line)
plt.title("Red Line")
plt.axis("off")
plt.subplot(1, 3, 2)
plt.imshow(yellow_line)
plt.title("Yellow Line")
plt.axis("off")
plt.subplot(1, 3, 3)
plt.imshow(blue_line)
plt.title("Blue Line")
plt.axis("off")
plt.show()
