伸縮自如的Flask [day 26] Flask with ML

flask scikit-learn

wilsonsujames 2021-06-09 14:37:40 ‧ 2048 瀏覽

分享至

github:
https://github.com/wilsonsujames/ML_on_web

資料集:
https://www.kaggle.com/shwetabh123/mall-customers

這次是用機器學習中的kmeans演算法作範例，如何將AI執行出來預測的結果視覺化。

首先是模型的訓練 kmean_train.py :


#load 進資料集
dataset = pd.read_csv('./Mall_Customers.csv')

#可以切分訓練集
X= dataset.iloc[:150, [3,4]].values

#從分2類到分11類來評估最佳分群數
silhouette_avg = []
for i in range(2,11):
    try:
        kmeans = KMeans(n_clusters= i, init='k-means++').fit(X)
        silhouette_avg.append(silhouette_score(X, kmeans.labels_)) 
    except:
        pass

indArr,peak_heightsDict=find_peaks(silhouette_avg)

# 在silhouette_avg這個list中，最高點index為三，最佳分群數為5
RowNumCategories=indArr[0]+2

# 模型預測及畫圖
kmeansmodel = KMeans(n_clusters= RowNumCategories, init='k-means++')
y_kmeans= kmeansmodel.fit_predict(X)

# 利用joblib 來將模型儲存
dump(kmeansmodel, 'kmean.joblib')

在dashboard.html的部分，放置了圖片及圖片來源:

<div class="container">
  <img src="{{url_for('plot_png')}}" alt="my plot">
</div>

在server.py的部分，由plot_png這個function來提供圖片:

@app.route('/plot')
def plot_png():

    dataset = pd.read_csv('./Mall_Customers.csv')
    # X= dataset.iloc[151:202, [3,4]].values
    X= dataset.iloc[:, [3,4]].values
    kmean_clf = load('kmean.joblib') 
    y_kmeans= kmean_clf.fit_predict(X)

    Kmeansfig = Figure()
    axis = Kmeansfig.add_subplot(1, 1, 1)
    axis.grid(color='lightgrey')
    axis.scatter(X[y_kmeans == 0, 0], X[y_kmeans == 0, 1], s = 100, c = 'red', label = ' 1')
    axis.scatter(X[y_kmeans == 1, 0], X[y_kmeans == 1, 1], s = 100, c = 'blue', label = '2')
    axis.scatter(X[y_kmeans == 2, 0], X[y_kmeans == 2, 1], s = 100, c = 'green', label = '3')
    axis.scatter(X[y_kmeans == 3, 0], X[y_kmeans == 3, 1], s = 100, c = 'cyan', label = ' 4')
    axis.scatter(X[y_kmeans == 4, 0], X[y_kmeans == 4, 1], s = 100, c = 'magenta', label = '5')
    axis.scatter(kmean_clf.cluster_centers_[:, 0], 
    kmean_clf.cluster_centers_[:, 1], s = 300, c = 'yellow', label = 'Centroids')
    axis.set_title("K-means demo")
    axis.legend()

    output = io.BytesIO()
    FigureCanvas(Kmeansfig).print_png(output)
    
    return Response(output.getvalue(), mimetype="image/png")

使用joblib來load進模型物件，預測給定的pandas資料，將預測的5群及群心分別上不同顏色，將圖片結果呈現在網頁中。

結果呈現: