Vizualizatsiya: Matplotlib va Seaborn

uv add matplotlib seaborn

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

# 1. "pyplot" uslubi — tez, lekin global holatga tayanadi
plt.plot([1, 2, 3], [1, 4, 9])
plt.title("Kvadratlar")
plt.show()
 
# 2. Obyektga yo'naltirilgan uslub — ✅ tavsiya etiladi
fig, ax = plt.subplots(figsize=(6, 4))
ax.plot([1, 2, 3], [1, 4, 9])
ax.set_title("Kvadratlar")
fig.savefig("squares.png", dpi=150)

fig, ax = plt.subplots()
 
ax.plot(x, y, label="model A", linestyle="--", marker="o")   # chiziq — trend, vaqt
ax.scatter(x, y, s=10, alpha=0.5, c=labels)                  # nuqtalar — ikki o'zgaruvchi munosabati
ax.hist(values, bins=40)                                     # gistogramma — taqsimot
ax.bar(categories, heights)                                  # ustunlar — kategoriyalarni solishtirish
ax.barh(categories, widths)                                  # gorizontal — uzun nomlar uchun
ax.boxplot([group1, group2])                                 # quti — taqsimotlarni solishtirish
ax.imshow(matrix, cmap="viridis")                            # matritsa / rasm
ax.errorbar(x, means, yerr=stds, capsize=3)                  # xato chegaralari bilan
ax.fill_between(x, lower, upper, alpha=0.2)                  # oraliq (ishonch intervali)

ax.set(title="Sarlavha", xlabel="X o'qi", ylabel="Y o'qi")   # bir nechtasini birdan
ax.set_xlim(0, 100); ax.set_ylim(bottom=0)
ax.set_yscale("log")                                         # logarifmik shkala
ax.legend(loc="upper right")                                 # label= berilganlar uchun
ax.grid(alpha=0.3)
ax.axhline(0.9, color="gray", linestyle=":")                 # gorizontal chiziq (maqsad, baseline)
ax.annotate("cho'qqi", xy=(x0, y0), xytext=(x0 + 5, y0 + 1), arrowprops=dict(arrowstyle="->"))

df = pd.read_csv("orders.csv", parse_dates=["date"])
df["revenue"] = df["price"] * df["qty"]
 
fig, axes = plt.subplots(2, 2, figsize=(11, 7.5), layout="constrained")   # axes — (2, 2) NumPy massiv
 
axes[0, 0].hist(df["price"] / 1000, bins=40)
axes[0, 0].set(title="Narx taqsimoti", xlabel="ming so'm", ylabel="buyurtmalar")
 
axes[0, 1].hist(np.log10(df["price"]), bins=40)
axes[0, 1].set(title="Narx (log10 shkalada)", xlabel="log10(narx)")
 
(df.groupby("city")["revenue"].sum() / 1e6).sort_values().plot.barh(ax=axes[1, 0])
axes[1, 0].set(title="Shaharlar bo'yicha tushum", xlabel="mln so'm", ylabel="")
 
monthly = df.set_index("date").resample("ME")["revenue"].sum() / 1e6
axes[1, 1].bar(monthly.index.strftime("%Y-%m"), monthly.values)
axes[1, 1].set(title="Oylik tushum", ylabel="mln so'm")
axes[1, 1].tick_params(axis="x", rotation=45)
 
for ax in axes.flat:
    ax.grid(alpha=0.3)
fig.suptitle("Buyurtmalar: EDA")
fig.savefig("eda.png", dpi=120)

rng = np.random.default_rng(0)
epochs = np.arange(1, 31)
train_loss = 2.0 * np.exp(-epochs / 8) + 0.15 + rng.normal(0, 0.02, 30)
val_loss = 2.0 * np.exp(-epochs / 8) + 0.3 + np.maximum(0, (epochs - 18) * 0.02) + rng.normal(0, 0.03, 30)
 
fig, ax = plt.subplots(figsize=(8, 4.5), layout="constrained")
ax.plot(epochs, train_loss, label="train")
ax.plot(epochs, val_loss, label="validation")
 
best = int(np.argmin(val_loss)) + 1
ax.axvline(best, ls="--", c="gray")
ax.annotate(f"eng yaxshi: {best}-epoch", xy=(best, val_loss.min()),
            xytext=(best + 2, val_loss.min() + 0.4), arrowprops=dict(arrowstyle="->"))
ax.set(xlabel="Epoch", ylabel="Loss", title="O'qitish egri chiziqlari")
ax.legend()
ax.grid(alpha=0.3)
fig.savefig("curves.png", dpi=150)

df["price"].plot.hist(bins=40)
df.groupby("category")["revenue"].sum().sort_values().plot.barh()
monthly.plot(marker="o")
df.plot.scatter(x="price", y="rating", alpha=0.3)
 
ax = df.groupby("city")["price"].median().plot.bar(rot=0)   # Axes qaytaradi — bezash mumkin
ax.set_ylabel("mediana narx")

import seaborn as sns
 
sns.set_theme(style="whitegrid")
 
fig, axes = plt.subplots(1, 3, figsize=(15, 4), layout="constrained")
sns.histplot(data=df, x="price", hue="category", log_scale=True, element="step", ax=axes[0])
sns.boxplot(data=df, x="city", y="price", log_scale=True, ax=axes[1])
sns.barplot(data=df, x="category", y="rating", errorbar=("ci", 95), ax=axes[2])   # o'rtacha + 95% interval

num = df[["price", "qty", "rating", "revenue"]]
sns.heatmap(num.corr(), annot=True, fmt=".2f", cmap="coolwarm", vmin=-1, vmax=1, square=True)

fig.savefig("plot.png", dpi=150, bbox_inches="tight")   # hisobot uchun PNG
fig.savefig("plot.svg")                                  # vektor — veb va maqolalar uchun
fig.savefig("plot.pdf")                                  # LaTeX maqolalar uchun
plt.close(fig)                                           # siklda ko'p grafik yaratsangiz — xotirani bo'shating

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
 
 
def eval_report(evals: pd.DataFrame, path: str) -> None:
    summary = evals.groupby("variant").agg(
        accuracy=("correct", "mean"),
        n=("correct", "size"),
        cost=("cost_usd", "mean"),
        p95=("latency_s", lambda s: s.quantile(0.95)),
    )
    se = np.sqrt(summary["accuracy"] * (1 - summary["accuracy"]) / summary["n"])   # standart xato
 
    fig, axes = plt.subplots(1, 3, figsize=(15, 4.2), layout="constrained")
 
    order = summary["accuracy"].sort_values().index
    axes[0].barh(order, summary.loc[order, "accuracy"], xerr=1.96 * se[order], capsize=4)
    axes[0].set(title="Aniqlik (95% interval)", xlim=(0, 1))
 
    for variant, g in evals.groupby("variant"):
        axes[1].hist(g["latency_s"], bins=30, alpha=0.5, label=variant)
    axes[1].set(title="Kechikish taqsimoti", xlabel="s")
    axes[1].legend()
 
    axes[2].scatter(summary["cost"] * 1000, summary["accuracy"], s=80)
    for name, row in summary.iterrows():
        axes[2].annotate(name, (row["cost"] * 1000, row["accuracy"]), xytext=(5, 5), textcoords="offset points")
    axes[2].set(title="Narx va aniqlik", xlabel="$ / 1000 so'rov", ylabel="aniqlik")
 
    for ax in axes:
        ax.grid(alpha=0.3)
    fig.savefig(path, dpi=150)
    plt.close(fig)