# Music Analysis Model

In this page we explain the model we have followed to mathematically analyze what makes music more popular with parameters like "danceability, mood, liveness" for example.

```
import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from sklearn import datasets, linear_model
from sklearn.decomposition import PCA
from sklearn.preprocessing import MinMaxScaler
from sklearn.manifold import TSNE

%matplotlib inline

data_frame = pd.read_csv("../input/data.csv")
data_frame = data_frame.drop("Unnamed: 0", axis="columns")
data_frame.head()
```

For our first trick we'll create a scatterplot based on the songs' valence and danceability values. In addition, we will also use linear regression to find their correlation.

```
x = data_frame["danceability"].values
y = data_frame["valence"].values

x = x.reshape(x.shape[0], 1)
y = y.reshape(y.shape[0], 1)

regr = linear_model.LinearRegression()
regr.fit(x, y)

fig = plt.figure(figsize=(6, 6))
fig.suptitle("Correlation between danceability and song mood")

ax = plt.subplot(1, 1, 1)
ax.scatter(x, y, alpha=0.5)
ax.plot(x, regr.predict(x), color="red", linewidth=3)
plt.xticks(())
plt.yticks(())

ax.xaxis.set_major_locator(ticker.MultipleLocator(0.1))
ax.xaxis.set_minor_locator(ticker.MultipleLocator(0.02))

ax.yaxis.set_major_locator(ticker.MultipleLocator(0.1))
ax.yaxis.set_minor_locator(ticker.MultipleLocator(0.02))

plt.xlabel("danceability")
plt.ylabel("valence")

plt.show()
```

<figure><img src="/files/ZIJslRCYNXNeNyZmnfnw" alt=""><figcaption></figcaption></figure>

Now to create some histograms. The plot on the left illustrates the distribution of songs based on their energy levels, whereas the one on the right is a "heatmap" (histogram in two dimensions) that illustrates the number of songs found at all values of valence and danceability.

```
x = "danceability"
y = "valence"

fig, (ax1, ax2) = plt.subplots(1, 2, sharey=False, sharex=False, figsize=(10, 5))
fig.suptitle("Histograms")
h = ax2.hist2d(data_frame[x], data_frame[y], bins=20)
ax1.hist(data_frame["energy"])

ax2.set_xlabel(x)
ax2.set_ylabel(y)

ax1.set_xlabel("energy")

plt.colorbar(h[3], ax=ax2)

plt.show()
```

<figure><img src="/files/dwM9Rwc1RKWKwUM6WTyu" alt=""><figcaption></figcaption></figure>

## This is where the fun begins[¶](https://www.kaggleusercontent.com/kf/1920358/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..o68Mrv90MnVbuTmYUHkBbQ.9kFIOiTvmRYl-RJNrEUOXRf9PEBrDQ2Hnlwo-d2cv1eG-bg7_FKT4KeStA5AetAnUEsATKE482uE3zQimPGhrqhj-MJYrPLs6NwHK4mo9N9yWR7SuBqAghv-3SXH8Jlm8rTVbXIE3jXIuS1KpET4a_vi3_U3UOzXsp1lMif7yDMsOUebGF5gzOPjsXeIHzfi_IYU6qGYwaOPeJo7XveeMTnzfsJyP9i1V1ZwAEaAXm5-OhPvHaWroRcnW18zTedwKVAf9MyZH0R-VzYXmFIHQbKLCNGL4OKcdMzQT1u0CgQnw2D5o5uZ--Vd7aIEaNT48ZS5WUFkXLU-H0jA4fHs5E09n-5ATm837b83Eox7hbitu1H0JxJ48G9mcbRFJ8edMOzDv647vlTCINp-5kXrUOGZUCmolsrHgU6NnUb_sqD81k8q8xE4cjng_uHY7Oq8ljwCPygiARNSEqUkRtdcupFvE0hmUOyTe0IUVOSCoWhH8T3Y7dMwtcLxecIIC79GJA1YbAQhKR4AHpXN0I_5dVLvbMIqWYU-sQfr1RZuUZq1MW4o52grPYyjyeLQcvpRJJnj6AtBjpDLtAsEtoPMd6HWC7kVAy8qAFXh3NBU-v2GKnc9OpRyiXHBmoDJwkPL3t25aGC61qxHY8Kpm4AGWg.XRdBPHGt2Oli6hXaFbD7EQ/__results__.html?sharingControls=true#This-is-where-the-fun-begins) <a href="#this-is-where-the-fun-begins" id="this-is-where-the-fun-begins"></a>

Next, we generate a list of "chosen" traits and use principal component analysis to reduce the dimensions of that list to 3, effectively creating a matrix suitable for generating the following 3-dimensional plot. The lesser the distance between any two songs, the larger the similarities in their traits.

```
chosen = ["energy", "liveness", "tempo", "valence", "loudness", "speechiness", "acousticness", "danceability", "instrumentalness"]
text1 = data_frame["artist"] + " - " + data_frame["song_title"]
text2 = text1.values

# X = data_frame.drop(droppable, axis=1).values
X = data_frame[chosen].values
y = data_frame["danceability"].values

min_max_scaler = MinMaxScaler()
X = min_max_scaler.fit_transform(X)

pca = PCA(n_components=3)
pca.fit(X)

X = pca.transform(X)

import plotly.offline as py
py.init_notebook_mode(connected=True)
import plotly.graph_objs as go
import plotly.tools as tls

trace = go.Scatter3d(
    x=X[:,0],
    y=X[:,1],
    z=X[:,2],
    text=text2,
    mode="markers",
    marker=dict(
        size=8,
        color=y
    )
)

fig = go.Figure(data=[trace])
py.iplot(fig, filename="test-graph")
```

In another attempt, we use a smaller "chosen" list and PCA to generate a two-dimensional graph.This specific graph was chosen due to its seemingly spot-on axes naming.<br>

<figure><img src="/files/M0WSAQlab6BZxrjV6QzN" alt=""><figcaption></figcaption></figure>

In another attempt, we use a smaller "chosen" list and PCA to generate a two-dimensional graph.This specific graph was chosen due to its seemingly spot-on axes naming.

```
chosen = ["energy", "liveness", "tempo", "valence"]
text1 = data_frame["artist"] + " - " + data_frame["song_title"]
text2 = text1.values

# X = data_frame.drop(droppable, axis=1).values
X = data_frame[chosen].values
y = data_frame["loudness"].values

min_max_scaler = MinMaxScaler()
X = min_max_scaler.fit_transform(X)

pca = PCA(n_components=2)
pca.fit(X)

X = pca.transform(X)

fig = {
    "data": [
        {
            "x": X[:, 0],
            "y": X[:, 1],
            "text": text2,
            "mode": "markers",
            "marker": {"size": "8", "color": y}
        }
    ],
    "layout": {
        "xaxis": {"title": "How hard is this to dance to?"},
        "yaxis": {"title": "How metal is this?"}
    }
}

py.iplot(fig, filename="test-graph2")
```

<figure><img src="/files/SBoWOO6o5bV93p4wOGni" alt=""><figcaption></figcaption></figure>

Last but not least we generate a similar graph using t-SNE and yet another "chosen" list instead.

```
import time

chosen = ["energy", "liveness", "tempo", "valence", "loudness",
          "speechiness", "acousticness", "danceability", "instrumentalness"]

X = data_frame[chosen].values
y = data_frame["loudness"].values

min_max_scaler = MinMaxScaler()
X = min_max_scaler.fit_transform(X)

time_start = time.time()
tsne = TSNE(n_components=2, verbose=1, perplexity=40, n_iter=300)
tsne_results = tsne.fit_transform(X)

print('t-SNE done! Time elapsed: {} seconds'.format(time.time()-time_start))

fig = {
    "data": [
        {
            "x": tsne_results[:, 0],
            "y": tsne_results[:, 1],
            "text": text2,
            "mode": "markers",
            "marker": {"size": "8", "color": y}
        }
    ],
    "layout": {
        "xaxis": {"title": "x-tsne"},
        "yaxis": {"title": "y-tsne"}
    }
}

py.iplot(fig, filename="test-graph2")
```

```
[t-SNE] Computing 121 nearest neighbors...
[t-SNE] Indexed 2017 samples in 0.002s...
[t-SNE] Computed neighbors for 2017 samples in 0.155s...
[t-SNE] Computed conditional probabilities for sample 1000 / 2017
[t-SNE] Computed conditional probabilities for sample 2000 / 2017
[t-SNE] Computed conditional probabilities for sample 2017 / 2017
[t-SNE] Mean sigma: 0.145413
[t-SNE] KL divergence after 250 iterations with early exaggeration: 71.526894
[t-SNE] Error after 300 iterations: 1.485660
t-SNE done! Time elapsed: 16.126283168792725 seconds
```

<figure><img src="/files/p2KcC83WAmXTFK6HCsVX" alt=""><figcaption></figcaption></figure>

\ <br>


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://docs.stereoai.app/tech-area/music-analysis-model.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.